[3.11] gh-94816: Improve coverage of decode_linetable (GH-94853)

[miss-islington]

This makes calls to co_lnotab to exercise this code, as well
as generating synthetically large code to exercise the corner
cases where line numbers need multiple bytes.

Automerge-Triggered-By: GH:brandtbucher
(cherry picked from commit 20b9d2a)

Co-authored-by: Michael Droettboom mdboom@gmail.com

• Issue: [coverage] codeobject.c: decode_linetable #94816

Automerge-Triggered-By: GH:brandtbucher

[miss-islington]

Status check is done, and it's a success ✅ .

[brandtbucher]

Hm, this backport revealed a weakness in the test: every time we change the number of instructions (including caches) in any of these code objects, these tests will need to be updated. I don't even think that just checking the length is future-proof?

[miss-islington]

@mdboom: Status check is done, and it's a success ❌ .

[mdboom]

Yikes. Yes, perhaps we revert this. My main concern for coverage here was that when there are 4-byte line numbers that things still make sense / don't crash. I think as long as we can assert that somehow, we can remove the doctest part of this entirely.

EDIT: I confused this with another coverage issue. In this case, we just don't call decode_linetable at all.

[iritkatriel]

@markshannon has created #90916, and I guess the same applies to test_code as well.

We should have a test for each of the features, but not a large number of test cases that all test many things.

[mdboom]

We should close this (I don't have permissions) and I filed a PR to revert this on main: #94982