make mypy 0.730 pass cleanly again

this avoid duplication in the Dockerfile and also allows override from the
command-line

This was probably broken around a428cd8c and
caused SLF4J logging configuration to be ignored, falling back to crazye DEBUG
logging in the MPH step.

it is possible we want to actually define it, as it is in swh-py-template,
but there is a more general issue of how to keep that tox.ini in sync with
all other repos --- which is postponed for now

leading to significant performance improvements

- int->pid maps can be fully written, even from scratch, as MutableMapping

- pid->int maps can only be updated for already existing keys, due to the joint
  effect of binary search and fixed sized-ness

sequential ordering is only marginally useful, as generally we do not have the
list of PIDs ordered by sequential node ID available *before* creating this map

Python part only for now, the generation on the Java side will follow

thanks @haltode for noticing!

this is consistent with authorship practices in the project

haltode is still credited in the CONTRIBUTORS file + git history

using astyle in batch

Remove @version and @since javadoc unnecessary attributes.

Since 101c50f2, the `visited` array is not reset before each
Traversal methods but instead only once in class constructor. This
change broke unit tests which were using multiple times the same
Endpoint.

Unit tests were *sometimes* failing depending on the server startup
time.

This was creating a fixed overhead in the benchmarks. The bit vector is
by default initialized with false when instantiated.

Move CSV log creation out of timeEndpoint method, because some
benchmarks time multiple endpoints resulting in overwriting previous
logs.