[Kinflou]

Core part of #97. Stacked on #104 (std's validators need string); GitHub retargets this to master when #104 merges.

Syntax

validator StringBounds for string { ... }
validator Range for u8, u16, u32 { ... }

Checks (all build errors)

Code	When
validation::validator-missing-target	a validator has no for. The grammar keeps for optional only so this can name what's missing, with a fix-it hint.
validation::validator-target-mismatch	at the field: "validator 'StringBounds' applies to string, not s32". Exact match after resolving aliases ("not Count (u32)"). A validator on string[] checks the array, not each element. optional doesn't matter.
validation::validator-accessor-unsupported	value.length on a target without a length, or a value accessor other than length / name

Imported validators and aliases

These were skipped entirely before: the declaring schema isn't among the importing schema's units. validate_with(units, &dyn ImportLookup) lets the caller answer, following the same split as size.rs's SizeLookup:

• Build: ProjectImports follows the schema's use statements through the ProjectContext, std included.
• validate(): knows no imports, so behaviour there is unchanged.
• Language server: implements the trait in its own PR, using the shared declared_validator_targets / declared_alias_target.

Also

• for is a keyword (vocabulary: 25 keywords).
• std's StringBounds is now for string.
• Existing validator fixtures gain for string.
• Tests cover a missing target; mismatches on a local validator, through an alias, on an array, and with a list of targets; optional and aliased fields that match; both accessor errors; and the cross-schema cases (std's StringBounds, plus an alias and a validator from another schema in the same package).