Import JSON Schema as IDL #2

What
Add a flatc importer for JSON Schema that reconstructs the same IR you’d get from parsing the original .fbs, when the JSON Schema was produced by FlatBuffers’ exporter.

Why
Complete the round-trip (IDL → JSON Schema → IR) to reduce friction for teams that start with JSON Schema and to enable deterministic CI tests.

How
• New mode: flatc --schema-in path/to/schema.json [--lang ...]
• Prefer x-flatbuffers-* hints emitted by the exporter.
• Map JSON Schema → FlatBuffers:
• type: object → table; x-flatbuffers-struct: true → struct
• type: array → vector
• integer/string enums → enum (respect x-flatbuffers-enum-base)
• oneOf (+ discriminator/x-flatbuffers-union) → union
• $defs/$ref → named types/refs; required → (required); default → field default
• Map-like additionalProperties with x-flatbuffers-map: true
• Type inference from minimum/maximum:
• Choose narrowest signed/unsigned scalar (byte…ulong) covering bounds
• number defaults to double unless x-flatbuffers-float:"float"
• Clear diagnostics for unmappable features (e.g., complex allOf, patternProperties).

Done when
• Importing JSON Schemas generated by flatc yields IR ≡ importing the source .fbs.
• Minimal-width integer/unsigned types are inferred correctly from bounds.
• Unions/enums/structs/tables and FB attributes round-trip exactly.
• Helpful errors for unsupported constructs.
• Tests: golden round-trip corpus, numeric-inference property tests, unions/maps cases.