The rounding algorithm wasn't specified

Serializing as ISO8601 makes the hash brittle, because the database may
change the timezone silently and/or lose precision in the microseconds.

As we do not need precise timestamp, using an integer is good enough,
and is consistant with the git format.

The manifest also does not need to contain a timezone, as it only
represents the timezone of the system that fetched this metadata,
which is useless data.

So that they can be properly deduplicated and referenced.

This will be used to compute an intrisic identifier for RawExtrinsicMetadata;
which can be used for deduplication and refering to it like any other sha1_git
instead of needed to use a tuple of its fields.

Other packages don't use it anymore.

This is a useful shorthand when generating SWHIDs in tests.

SWHID is deprecated; and CoreSWHID does not support qualifiers at all,
so RawExtrinsicMetadata no longer needs to check there are no qualifiers.

ExtendedSWHID can identify either a software artifact or an origin,
so we no longer need Union[SWHID, str].

Therefore, we no longer need the 'type' attribute, as it was only
used to tell when the target is a SWHID and when it's an origin URL.

It can be handy as a shortcut to build SWHID objects.

* Quote/unquote path
* Fix line parsing and serializing to properly handle None
* Fix error raised by check_visit/check_anchor

They were all very similar and only differ in what 'edge' cases they accept

by making them all derive from an abstract class.

Following the discussion on T3034, we decided to replace SWHID with
two or three classes:

* QualifiedSWHID to replace the existing SWHID (standard types + qualifiers)
* CoreSWHID, for "core SWHID" only (standard types + no qualifiers)
* ExtendedSWHID for internal use in Software Heritage (extra types + no qualifiers)

This commit adds the last one.

It also removes "ori" as a valid object type for CoreSWHID and
QualifiedSWHID, as it now only belongs in ExtendedSWHID.

It is cleaner, avoids warnings, and will be needed when introducing
ExtendedSWHID in a future commit.

And store their parsed values (CoreSWHID, tuple of ints, etc.) instead of string.

Following the discussion on T3034, we decided to replace SWHID with
two or three classes:

* QualifiedSWHID to replace the existing SWHID (standard types + qualifiers)
* CoreSWHID, for "core SWHID" only (standard types + no qualifiers)
* ExtendedSWHID for internal use in Software Heritage (extra types + no qualifiers)

This commit adds the second one

Following the discussion on T3034, we decided to replace SWHID with
two or three classes:

* QualifiedSWHID to replace the existing SWHID (standard types + qualifiers)
* CoreSWHID, for "core SWHID" only (standard types + no qualifiers)
* ExtendedSWHID for internal use in Software Heritage (extra types + no qualifiers)

Since migrating from SWHID will break existing code, this commit uses
the opportunity to modernize it a little, ie.:

* `keyword`-only constructor, to get rid of the hacky default values for
  `object_type` and `object_id`
* enum instead of strings for the object type
* `bytes` instead of an hex string for the object id
* rename `metadata` to `qualifiers`

Depending on the order in which tests are run, these tests may insert
lambdas with mocked blake2 functions in their closure to be inserted in
hashutil._blake2_hash_cache; causing all future tests to fail.

While this does not happen with the default order of tests, it does when
using pytest-xdist.

so it's kept up to date when evolutions are made in the model and thus
preventing swh-journal and swh-model to be unecessarly coupled.

Related to T2970.

Related to T2961

The optional nature of the type will allow to avoid migrating the
(db) data model right now then we can have this type field in kafka
messages in the origin-visit-status topic

Related to T2443

They were mixed in with snapshot tests.

test_identifiers: Make sure that {directory,revision,release,snapshot}_identifier() doesn't just return a value from the dict.

For example, before this commit, you could replace the code of
revision_identifier() with this:

def release_identifier(release):
    return release.get("id", b"")

and all tests would still pass.

Before this change there was a lot of overlap between parse_swhid() and the
attrs-based validators in the SWHID class. Also, the validation implementation
in parse_swhid() was done by hand.

With this change the coarse-grained validation done by parse_swhid() is now
delegated to a regex. The semantic validation of SWHIDs is left to attrs
validators. The regex is also exposed as a module attribute, to be used by
client code that want to syntactically validate SWHIDs without necessarily
instantiate SWHID classes (we have several other modules doing that already,
and they are using slightly different hand-made regexs, which isn't great).

As part of this change we also clean up the use of ValidationError exceptions,
systematically passing the problematic parts of SWHID as arguments, and uniform
error messages.

This change also brings some speed up in SWHID parsing. On a benchmark parsing
~30 M valid SWHIDs, the previous implementation took ~3:06 minutes, the new one
~2:50 minutes, or a ~9% speedup.

Closes T2788

Unfortunately, sphinx (actually, autodoc) only picks up attributes if
they fall in any of these cases:

1. are enum variants
2. are in slots
3. are in __dict__
4. have an annotation
5. are found using its custom parser

(see get_object_members in sphinx/ext/autodoc/importer.py)

In theory, option 5 should work for us; unfortunately, autodoc only
asks the parser the list of members with a comment.
And it's not easy to adapt it to ask the parser for all members,
because said parser (sphinx/pycode/parser.py) does not return the class
qualname (aka. namespace) for members without comments.

So, as I don't want to change the interface of sphinx.pycode.parser,
this commit switches to relying on option 3, by adding __slots__ for
all attr classes.

Additionally, this might have some performance/memory improvement
(though I did not check) and will further avoid mutation of these
objects.

All reverse dependencies have been updated to avoid using it now, so it can now
be removed, paving the way to recycle it into an intrinsic identifier.

Related to T2729

So parse_swhid raises a ValidationError when that is detected.

Related to T2769

Related to T2769

Related to T2769

Related to T2769

This avoids throwing the deprecation warning when id and target are present and
have the same value, which makes a from_dict / to_dict round-trip throw no
deprecation warnings.