fix: preserve Fn::FindInMap with unresolved Ref keys in PARTIAL mode

[polyomino24]

Which issue(s) does this change fix?

#9004

Why is this change necessary?

sam build (1.160.0) fails with Fn::FindInMap layout is incorrect for any AWS::LanguageExtensions template that uses Fn::FindInMap with a !Ref Param as one of the keys, when no value is supplied for that parameter at build time (no --parameter-overrides, no Default). The same template built successfully with 1.159.1 — this is a regression introduced together with the new cfn_language_extensions module.

Root cause: FnRefResolver correctly preserves an unresolved {"Ref": "ENV"} dict in ResolutionMode.PARTIAL (the default for SAM CLI integration), but FnFindInMapResolver then validated the resolved keys with isinstance(..., str) and raised InvalidTemplateException unconditionally, regardless of resolution mode.

How does it address the issue?

In samcli/lib/cfn_language_extensions/resolvers/fn_find_in_map.py, mirror the FnRefResolver PARTIAL-mode pattern: when any resolved FindInMap key arrives as a dict (an unresolved nested intrinsic) and resolution_mode == PARTIAL, return the original Fn::FindInMap expression so CloudFormation can resolve it at deploy time. Non-dict, non-str keys (e.g. an int literal) remain a layout error in either mode, preserving existing behavior for genuine template mistakes.

What side effects does this change have?

• 3 Kotlin-compatibility tests in test_kotlin_compatibility.py previously asserted that Fn::FindInMap with Ref/GetAtt to a resource raises in default (PARTIAL) mode. That assertion captured the Kotlin reference's strict semantics, which conflict with SAM CLI's documented PARTIAL intent ("preserve unresolvable references in the output"). Those templates were moved to a new ERROR_TEMPLATES_FULL_MODE_ONLY list and are now asserted under ResolutionMode.FULL, keeping Kotlin parity for the strict callers while letting PARTIAL mode pass the expression through to CloudFormation. The remaining fnFindInMapWithUnsupportedFunctionFnSub template is unaffected (it already errored via a different code path).
• No public API changes. No changes to FULL mode behavior.

Mandatory Checklist

PRs will only be reviewed after checklist is complete

• Review the generative AI contribution guidelines
• Add input/output type hints to new functions/methods
• Write design document if needed (Do I need to write a design document?)
• Write/update unit tests
• Write/update integration tests
• Write/update functional tests if needed
• `make pr` passes
• `make update-reproducible-reqs` if dependencies were changed
• Write documentation

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.