fix: defer fetch request when repo is not yet initialized during GitProvider init

[waterWang]

Fixes #70081

Summary

During GitProvider.__init__(), fetch_request_check() is called (line 531) before the provider subclass has had a chance to call init_remote(). If a fetch_request file happens to exist in the salt working dir at this point (e.g. left over from a prior run), the code path:

fetch_request_check() -> self.fetch() -> self._fetch()

immediately accesses self.repo.remotes[0] (Pygit2 _fetch, and GitPython _fetch both do this). But self.repo is only set later, inside init_remote(), which the subclasses call after super().__init__() returns. The result is an intermittent AttributeError: 'Pygit2' object has no attribute 'repo' (or 'GitPython') that makes git_pillar/gitfs highstates fail every few runs.

Fix

In fetch_request_check(), check getattr(self, "repo", None). If the repo is not yet initialized, return False without removing the fetch_request file — the file is preserved so the next call to fetch_request_check() (from checkout(), by which time self.repo exists) processes it.

• Applies to both Pygit2 and GitPython providers.
• GitCLI is unaffected (its _fetch() shells out to the git binary and never touches self.repo).

Tests

• tests/pytests/functional/utils/gitfs/test_gitfs.py
• tests/pytests/functional/utils/gitfs/test_pillar.py