refactor(linalg): Add missing overload to compute_gramian

[ValerianRey]

I thought we could remove it, but this overload is actually necessary. Otherwise we get a Pylance error when we don't match the other overloads.

[ValerianRey]

@PierreQuinton FYI (you were right about this in your PR).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/torchjd/_linalg/_gramian.py	100.00% <ø> (ø)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[PierreQuinton]

@ValerianRey Any reason why the CI passed on this? I would have expected mypy to be unhappy there. Is there something we can do to make it stricter? Or is it just that mypy doesn't check overloads at all?

[ValerianRey]

@ValerianRey Any reason why the CI passed on this? I would have expected mypy to be unhappy there. Is there something we can do to make it stricter? Or is it just that mypy doesn't check overloads at all?

The reason is that we don't currently ever use compute_gramian with Tensor and contracted_dims other than the default, -1 or 1, so we never fell into this case.

Also, mypy isn't checked in tests.