Fix intrinsic resolution; fix effect param address space propagation; more mir cleanup

[sbillig]

MIR was resolving any function in core/std named "mload", "mstore", etc as an intrinsic (including trait methods). Fixing this by tracking and checking func.is_extern (this should be tightened up more, to check a specific path) revealed that these methods weren't resolvable to concrete impls when used in free functions called from contracts, because the concrete effect provider type from the contract wasn't being propagated to the functions called within the contract.

So, some reworking of how we do things to make this tidy(er):

• We treat all effect params as implicit generic params with a trait bound. For an effect that resolves to a concrete type T, the implicit generic param is __effprov0: EffectRef<T>. This has been the mental model for a while, now we're just fully leaning into it.
• Each resolved effect arg now carries key_kind (type vs trait) and (for type effects) the instantiated target type, so MIR doesn’t need to resolve/normalize effect keys/types.
• MIR now computes effect-param address spaces from HIR’s provider-param index map and uses HIR metadata instead of MIR path resolution.
• Removed obsolete contract RootEffect folding in MIR
• Improved monomorphization root seeding: use effect_bindings() as the source of truth and only seed the target default root effect instance when it satisfies the trait (error otherwise for now).
• Propagate address-space provenance for by-ref trait-effect providers via monomorphization specialization (and fix trait-method receiver-space for Self-typed trait methods), so storage-backed providers generate sload.
• Reduce redundant monomorphizations: canonicalize receiver_space=Memory to “no specialization”, eliminating duplicate *_mem instances.

[chatgpt-codex-connector]

💡 Codex Review

fe/crates/hir/src/analysis/ty/ty_check/expr.rs

Lines 1268 to 1272 in 41c0f4a

	PathRes::Trait(trait_inst) => {
	let mut instantiation_args = Vec::with_capacity(1 + callable.generic_args().len());
	instantiation_args.push(provided_ty);
	instantiation_args.extend_from_slice(callable.generic_args());

Instantiate trait-effect generics with correct args

In resolve_effect_requirement, trait effects are instantiated with instantiation_args = [provided_ty] + callable.generic_args(). After this change, callable.generic_args() includes the implicit effect-provider parameters that were inserted ahead of user generics, so a key like uses (ctx: Ctx<T>) will be instantiated as Ctx<__effprov0> instead of Ctx<T>. This misbinds trait requirements and can cause valid calls to fail (or invalid ones to pass) whenever the effect key depends on function generics. You likely need to exclude the effect-provider params (or otherwise realign indices) when building instantiation_args for trait effects.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[sbillig]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: bc7a5a3fc5

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".