feat(core): add direct JSON byte paths for remote access

[thaodt]

Description

Adds a compatibility-preserving direct JSON byte path to AimDB's type-erased remote-access stack.

• Adds defaulted byte methods to RemoteSerialize, JsonCodec<T>, JsonRecordAccess and JsonBufferReader.
• Keeps existing custom codec and reader impls source-compatible through serde_json::Value fallbacks.
• Overrides the built-in serde path with direct serde_json::to_vec and serde_json::from_slice.
• Routes state-record record.get and subscription events through owned JSON Payloads without materializing an intermediate record serde_json::Value.
• Preserves the existing AimX v2 schema, authorization, lag handling, cancellation, backpressure, error behavior, ring drain cursor semantics and no-producer-override write safety.
• Keeps metadata, graph, query, admin, aggregate drain, record.set and write wrapper handling on the existing tree path. Inbound wire handling does not use hand-written JSON offset extraction.
• Adds semantic-equivalence, compatibility, allocation and criterion benchmark coverage.

Performance

The focused benchmarks measure the in-memory typed-record-to-AimX-envelope boundary. They do not include connector queues, scheduler wakeups, syscalls or physical wire time.

Validation

• make check
• cargo test -p aimdb-bench
• cargo bench -p aimdb-bench --bench b0_alloc_remote_json
• cargo bench -p aimdb-bench --bench b1_b2_remote_json

The full repository gate covers formatting, linting, host tests, valid feature combinations, no_std + alloc, Cortex-M, WASM, dependency policy, README drift, codegen drift and production simulation checks.

Related Issue

• Closes [FEAT] Add direct JSON byte paths for type-erased remote access #196

Checklist

• I have read the CONTRIBUTING.md document.
• My code follows the project's coding standards.
• I have added tests to cover my changes.
• All new and existing tests passed (make check).
• I have updated the documentation accordingly.

[lxsaah]

Lgtm! Non-blocking follow-up: converting the shared drain_values to try_recv_json_bytes would give record.drain the direct path via RawValue (already used by the envelope) and drop the re-serialization in record.get’s ring fallback as well. We should think of a follow-up issue here! Thanks!!