perf(bench): evaluate CBOR for remote-access IR

[thaodt]

Description

Investigates whether CBOR should replace JSON as the self-describing representation at AimDB's type-erased remote-access boundary.

This PR:

• adds deterministic numeric, nested and byte-heavy record fixtures
• compares the current JSON-tree path, direct JSON, CBOR-to-JSON transcoding, native CBOR trees and direct CBOR
• adds allocation, latency, throughput and request/write/subscription-event envelope benchmarks
• adds an isolated no_std Cortex-M fixture for comparing linked JSON and CBOR footprint
• documents the evidence and final decision in Design 046.

Result

• Direct JSON was 2.06x–2.49x faster to encode and 2.41x–3.54x faster to decode than the JSON-tree path.
• Native CBOR reduced ordinary structured payloads by about 26%, but did not outperform direct JSON CPU time.
• For the byte-heavy fixture, native CBOR reduced frame size by about 70% and was more than 12x faster in the in-memory envelope prototype.
• The comparable Cortex-M CBOR image required 9592 additional flash-like bytes (+38.5%) versus JSON.
• CBOR-to-JSON transcoding was consistently worse than serializing JSON directly.

The resulting decision is an evidence-backed no-go for replacing the current JSON representation or changing AimX v2 as part of this issue.

The recommended follow-up is a compatibility-preserving direct-JSON-bytes optimization. A native binary remote edge should only be considered separately when there is a concrete byte-heavy consumer, protocol negotiation, bounded decoding and a target memory budget.

CBOR remains isolated to the unpublished benchmark crate and standalone footprint fixture. This PR does not change any production API, connector behavior or wire protocol.

Related Issue

• Ref [FEAT] Evaluate CBOR as the self-describing IR in the type-erased core (alt to serde_json::Value) #156

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.

[thaodt]

@lxsaah open this PR for your review, its for evaluation, not intended for merge. Please check the PR's description for detail and run benches yourself to validate my result too.