fix(experimental elaborator): Fix duplicate resolve_type on self type and don't leak a trait impl's generics

[jfecher]

Description

Problem*

Resolves

Summary*

Fixes two issues:

• Use cached resolved_object_type instead of re-resolving the self type in declare_methods_on_struct and check_trait_impl_coherence. This is needed to prevent duplicate errors, and in the later case, since generics are not added to scope to be able to re-resolve the object type at that point.
• Fix a case when resolving trait impls that self.generics was not cleared afterward.

Additional Context

Down to 112 errors in the standard library now. The remaining errors are variants of:

• "Duplicate definitions of <global> found"
• "Duplicate definitions of <generic> found" - looks to be only one case where there is a generic directly on a trait method. fn hash<H>(self, state: &mut H) where H: Hasher;
• "Expression type is ambiguous" - from the two turbofish cases that haven't been merged in this branch yet
• "Expected type &mut _, found type H" - issues with auto-deref and trait methods possibly
• "No matching impl found for ..." - trait solver issues

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[jfecher]

To avoid fixing as many merge conflicts I'm going to keep this PR open and merge jf/elaborator-fixes5 through 8 into this one before merging this into master

[jfecher]

The diff in fixes7 got messed up so I'm going to merge master into both of these PRs anyway and merge this one to master.