func.yaml schema generator truncates fractional `minimum` values (e.g. 0.01 -> 0)

[aliok]

The func.yaml JSON schema is generated with github.com/alecthomas/jsonschema. Its number-keyword handling stores minimum/maximum as int (Type.Minimum int, reflect.go:60) and parses the tag value with strconv.Atoi — for both the jsonschema tag (numbericKeywords, reflect.go:614) and the jsonschema_extras tag (extraKeywords). A fractional minimum such as 0.01 fails to parse, the error is discarded, and the emitted schema gets minimum: 0.

As a result, schema-based validation is looser than the runtime ValidateScale. For example scale.kpa.target requires >= 0.01 at runtime, but the generated schema only enforces > 0 (via an exclusiveMinimum: true workaround), so a value like 0.001 passes the published schema/func_yaml-schema.json while ValidateScale rejects it at deploy time.

This is a pre-existing, generator-wide limitation affecting any fractional minimum (not specific to the scale fields): the library simply cannot express a non-integer minimum.

Possible fixes

• Post-process the generated schema in schema/generator/main.go to set fractional minima after reflection (targeted, but per-field).
• Switch to a jsonschema library / tag path that supports float minima, then regenerate schema/func_yaml-schema.json so schema validation matches ValidateScale.

Context

Surfaced by the Copilot review of #4063:
#4063 (comment)