Consider dropping pull request references from the git loader ingestion

The loader git currently filters out references considered not that interesting [1]:

• auto-merged github pull requests (reference names starting with refs/pulls and finishes with /merge)
• peeled refs (reference names finishing with ^{})

Nonetheless, the current loader git actually still load pull requests references and there can be a lot depending on the repository. See for example a recent snapshot on the torvalds/linux [2] repository.

We should consider whether that's still relevant to ingest those references.

The webapp already considers this noise and has been filtering them out from the browsing since v0.0.288 version [3]. So that tends toward ignoring them as well during the ingestion.

That should also alleviate other current considerations [4].

• [1] https://forge.softwareheritage.org/source/swh-loader-git/browse/master/swh/loader/git/utils.py$89-90

• [2] https://archive.softwareheritage.org/api/1/snapshot/c2847dfd741eae21606027cf29250d1ebcd63fb4/

• [3] rDWAPPScc652d5240

• [4] #3625 (closed)

---

Migrated from T3627 (view on Phabricator)