Record repo mapping entries for labels in module extension tags

[fmeum]

The mapping of an apparent name in a module extension tag's attr.label attribute to the corresponding canonical name has to be recorded in the lockfile so that instantiated repo rules don't reference the stale repos.

Fixes #25063

[meteorcloudy]

Thanks for the quick fix!

[Wyverald]

Thanks for the lightning-fast investigation and fix! A couple of things:

• Could you add an end-to-end regression test in bazel_lockfile_test.py?
• Just to confirm, back in Lockfile fails to be regenerated when use_repo_rule statements are added #25063 you said "We need to include the repo mapping entries we applied to each extension's attr.label tag attributes into the extension's usages hash." This is not what we're doing here, right? We're just recording the repo mapping entries that are used to parse the label attrs.

[fmeum]

• Could you add an end-to-end regression test in bazel_lockfile_test.py?

Done!

• Just to confirm, back in Lockfile fails to be regenerated when use_repo_rule statements are added #25063 you said "We need to include the repo mapping entries we applied to each extension's attr.label tag attributes into the extension's usages hash." This is not what we're doing here, right? We're just recording the repo mapping entries that are used to parse the label attrs.

That's correct. Once I understood the cause, my first thought was that this info would go into the usage hash together with all other tag data, but folding it into the existing repo mapping entries field turned out to be much easier.