Specialized subscript guards should require non-negative compact ints

[KRRT7]

The specialized int subscript fast paths currently assume a non-negative compact int, but that precondition is not expressed directly in the guard.

Make the guard explicit for the affected cases:

• list[i]
• tuple[i]
• str[i]
• list[i] = value

This aligns the guard with the existing specialization contract and removes redundant checks from the specialized bodies, without changing the fallback path for negative indices.

Linked PRs

• gh-153486: Factor non-negative compact int subscript guards #153487

[markshannon]

BINARY_OP_SUBSCR_LIST_INT doesn't require that the index is non-negative, so this will slowdown l[-1]
Negative indices are quite common, especially -1.

Perhaps, it would make more sense to handle negative indices for all these cases, like BINARY_OP_SUBSCR_LIST_INT already does:

index  += index < 0 ? len : 0;

before deopting if the index is still negative.

[KRRT7]

so this will slowdown l[-1]

yeah, I think i noticed the perf regression in the xml_etree_generate

Perhaps, it would make more sense to handle negative indices for all these cases

I'll look into it, thank you.