Go back to LBYL programming style (instead of EAFP) and several mypy fixes in fem.petsc and la.petsc

[francesco-ballarin]

• Works towards LBYL vs. EAFP (isinstance vs. try/catch) #3738 for fem.petsc and la.petsc
• Replace Iterable with Sequence in all public function signatures in fem.petsc and la.petsc. See Go back to LBYL programming style (instead of EAFP) and several mypy fixes in fem.petsc and la.petsc #3790 (comment) .
• Remove module level # mypy: ignore-errors in fem.petsc: this has uncovered several typing errors that are fixed here.

(replaces #3789 which was not branched off of main)

[schnellerhase]

I think there are quite a few more cases of the EAFP paradigm. In mesh.py:637-666, bcs.py:51-55, bcs.py:87-91, bcs.py:225-234, forms.py:358-362, function.py:144-147, function.py:475-484, function.py:602-613, function.py:628-631, function.py:683-690, fem/petsc.py (handled in #3785), io/utils.py:84-90, io/utils.py:97-104.

If not addressed here, we should keep the issue open.

[francesco-ballarin]

If not addressed here, we should keep the issue open.

I edited the PR description removing "Fixes".

[francesco-ballarin]

I now removed # mypy: ignore-errors, and the scope of the PR changed massively. Most of the changes are now related to mypy fixes which, as @jhale suspected, was basically untested due to the module level ignore.

I also moved from Iterable to Sequence in fem.petsc and la.petsc for all the reasons listed above.

I suggest giving the review of this PR a priority higher than #3787 #3794 . After this one is in, the other two PRs will finish up on fixing #3738, which was the original goal of this PR.

[schnellerhase]

We should maybe aim to activate warn_unused_ignores in the mypy settings to get notified when all these PETSc ignores are no longer necessary.

[francesco-ballarin]

We should maybe aim to activate warn_unused_ignores in the mypy settings to get notified when all these PETSc ignores are no longer necessary.

Agreed, but this will not happen in this PR.

[francesco-ballarin]

Principle agreed upon in previous discussions. Queueing this to merge to unblock other PRs (e.g. #3794).