test: reduce and improve tests in src/compile/types.rs

[github-actions]

Summary

Improves the test suite in src/compile/types.rs by strengthening weak tests and restoring genuine coverage that had been incorrectly identified as redundant.

Strengthened tests

Test	Action	Reason
test_supply_chain_feed_only_validates	Strengthened	Now also asserts that a combined feed + registry-with-connection config validates OK, exercising more of the validation logic
test_permissions_default	Rewritten	Now deserialises permissions: {} from YAML instead of calling Default::default() — tests the actual serde contract and guards against accidentally adding required fields

Retained tests (with corrections)

Test	Action	Reason
test_frontmatter_inlined_imports_false_explicit	Kept	Only test of the explicit inlined-imports: false override path — distinct from the default (absent key) path
test_cache_memory_bool_false	Kept with full assertions	assert!(cm.allowed_extensions().is_empty()) retained — verifies the Enabled variant's accessor returns no extensions, a distinct path from is_enabled()
test_lean_bool_false	Kept with full assertions	assert!(lean.toolchain().is_none()) retained — verifies the Enabled variant returns None; not covered by test_lean_with_toolchain which only exercises WithOptions

Test plan

• cargo test: all tests pass ✅
• cargo clippy --all-targets --all-features: no warnings or errors ✅

[jamesadevine]

@copilot The new assertions are welcome:

• the combined  feed  +  registry -with- service-connection   validate()  check in  test_supply_chain_feed_only_validates 
• switching  test_permissions_default  to deserialize  permissions: {}  (nicely guards against a future required field or a non- None  serde default)

However, three deletions drop real, non-redundant coverage:

1.  test_frontmatter_inlined_imports_false_explicit  (removed entirely) — this is the only test of the explicit  inlined-imports: false  override path. The remaining test covers the default/true case, so the false override is now untested. Please keep it.
2.  assert!(cm.allowed_extensions().is_empty())  (removed) — verifies that a disabled  cache-memory  config reports no allowed extensions. That's a distinct accessor path from  is_enabled() ; keeping the assertion costs nothing.
3.  assert!(lean.toolchain().is_none())  (removed) — this asserted the  Enabled  variant returns  None . The replacement comment says it's "covered by  test_lean_with_toolchain ", but that test exercises the  WithOptions  variant, not the  Enabled → None  path — so this is a genuine coverage gap, not a duplicate.

Could you restore these three? Trimming the truly redundant bits is fine — e.g. dropping  test_generate_stage_prefix -style single-word duplicates — but #1098 was just closed for removing the only coverage of a specific behavior, and items 1 and 3 here are the same pattern. Happy to merge once they're back in alongside the new assertions.

[Copilot]

@copilot The new assertions are welcome:

• the combined  feed  +  registry -with- service-connection   validate()  check in  test_supply_chain_feed_only_validates 
...

Restored all three in commit test: restore three dropped coverage items per review feedback:

1. test_frontmatter_inlined_imports_false_explicit — test is back, covering the explicit inlined-imports: false override path.
2. assert!(cm.allowed_extensions().is_empty()) — restored in test_cache_memory_bool_false.
3. assert!(lean.toolchain().is_none()) — restored in test_lean_bool_false (and the comment replacing it removed).