Fix potential fgIsBigOffset issue in optAssertionIsNonNull

[EgorBo]

Currently in optAssertionIsNonNull we start from peeling the offset from GenTree:

runtime/src/coreclr/jit/assertionprop.cpp

Lines 4618 to 4624 in 92741be

	bool Compiler::optAssertionIsNonNull(GenTree* op, ASSERT_VALARG_TP assertions)
	{
	if (op->OperIs(GT_ADD) && op->AsOp()->gtGetOp2()->IsCnsIntOrI() &&
	!fgIsBigOffset(op->AsOp()->gtGetOp2()->AsIntCon()->IconValue()))
	{
	op = op->AsOp()->gtGetOp1();
	}

Then, for Global-AP we peel it again on top of VN level:

runtime/src/coreclr/jit/assertionprop.cpp

Lines 4655 to 4657 in 92741be

	ValueNum vnBase = vn;
	target_ssize_t offset = 0;
	vnStore->PeelOffsets(&vnBase, &offset);

but this time we don't check the offset with fgIsBigOffset (in fact, we should check the sum of both offsets).

Instead, this PR only does VN-level peeling for global AP and uses a special gtPeelOffset helper for the local one.
Also, removed unsound use of gtEffectiveVal

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[Copilot]

Pull request overview

This PR updates CoreCLR JIT assertion propagation to avoid double-peeling offsets (tree-level + VN-level) in optAssertionIsNonNull, ensuring fgIsBigOffset is applied to the combined peeled offset and removing reliance on gtEffectiveVal for this analysis.

Changes:

• Switch global assertion-prop non-null checks to VN-only offset peeling with an explicit fgIsBigOffset guard.
• Switch local assertion-prop non-null checks to use gtPeelOffsets to compute the full summed offset before applying fgIsBigOffset.
• Remove the prior gtEffectiveVal-based restriction so global AP can reason about non-lclvar trees via VN.

[Copilot]

Pull request overview

Copilot reviewed 1 out of 1 changed files in this pull request and generated 1 comment.