Fix bug in QF.find_primitive_p_divisible_vector__next

[tornaria]

When the vector [0,...,0,1] is a zero mod p, it was missed from the iteration. Fix by check and return it if this is the case.

A doctest triggering the bug is included.

The second commit adds a return_matrix option for QF.find_p_neighbor_from_vec which is related and useful, similar to QF.is_globally_equivalent_to, etc.

📝 Checklist

• The title is concise, informative, and self-explanatory.
• The description explains in detail what this PR is about.
• I have created tests covering the changes.
• I have updated the documentation accordingly.

[tornaria]

Ping @yyyyx4 : if you find the time for a quick review, this is a very small PR that fixes a few annoying bugs in ternary quadratic forms. Most of the changed lines are new (simple) doctests for the bugs. Thanks.

[yyyyx4]

Your changes look good to me.

As a side remark (which is orthogonal to this patch), the API with __next here is really strange by Python standards: I suppose it should just be a generator, using yield.

[github-actions]

Documentation preview for this PR (built with commit 522edd1; changes) is ready! 🎉

[tornaria]

@vbraun I'd appreciate if this could be merged for 10.2, as it fixes two real annoying bugs and it's quite unobtrusive. I'm finishing a paper+code and it'd be nice if it works on unpatched 10.2.

[tornaria]

Quoting #36031 (comment):

Marking this as a blocker in the hope it can sneak into 10.2 with the other open blocker

[mkoeppe]

Yes, that's the documented mechanism to request that something be included in the release.