Formalize linetable parsing and propagate out-of-bounds errors

[noahbkim]

We've noticed a transient issue where py-spy will abort due to unmet assumptions about the contents of co_linetable (for example we've seen [231], which doesn't even satisfy the internal specification). To address this, I've patched the co_linetable parsing to propagate out-of-bounds errors so that the samples are simply ignored (and a warning is raised).

This change is not as minimal as it could be, for example it reorders a couple lines in python_interpreters.rs for clarity. If that's a problem, I'd be happy to go back and pare down such changes.

Taking a glance, I would guess this is the root cause for #735 and #569, although I have no explanation for why the co_linetable's appear invalid.

[noahbkim]

Updated the PR to include a fix for an off-by-one introduced by #635, trivially reproducible with the following:

# a.py
import b

# b.py
import time
time.sleep(10)

# Command
cargo run -F unwind --release -- record --native --rate 1 -- python3.12 a.py

[korniltsev]

This change is not as minimal as it could be, for example it reorders a couple lines in python_interpreters.rs for clarity. If that's a problem, I'd be happy to go back and pare down such changes.

Let's change only what's needed for the fix, so that it's easier to cherry-pick & revert & reason about

[Hnasar]

@benfred can approve this workflow so the build can run?

Otherwise this LGTM, and we've manually built it and confirmed it's fixed the issues we were seeing.