perf(sdk-go): reduce websocket order dispatch allocations

[namtran1812]

Summary

Reduce allocation overhead in the Go SDK WebSocket order-event dispatch path by replacing the generic map[string]json.RawMessage discriminator decode with a small typed envelope.

The envelope includes both e (event type) and E (event time) so encoding/json resolves the protocol's case-sensitive fields exactly while preserving the existing order-before-trade classification behavior.

Motivation

Private order updates may contain a top-level t trade ID. The dispatcher therefore needs to identify explicit order / orderUpdate events before applying the generic trade heuristic.

Previously, that classification decoded every candidate order frame into a map[string]json.RawMessage and then decoded the e field separately. This adds avoidable allocations on the WebSocket dispatch path.

Changes

• Replace the map-based order-event envelope with a typed e / E envelope.
• Preserve explicit order-event classification before trade heuristics.
• Extend the existing regression test to cover e, numeric E, and top-level t together.
• Add dispatch benchmarks for depth, trade, ticker, and order frames.

Benchmark

Apple M1, darwin/arm64:

	Before	After
Order dispatch	~5.19 µs/op	~3.96 µs/op
Allocated bytes	2,560 B/op	1,072 B/op
Allocations	44 allocs/op	20 allocs/op

This is approximately 58% fewer allocated bytes and 55% fewer allocations per order frame in the benchmark. Timing results are environment-specific.

Validation

• go test ./... -count=1
• go vet ./...
• go test -race ./websocket -count=1
• order/trade dispatch regression: 100 consecutive passes
• public/private stream separation regression: 50 consecutive passes
• git diff --check

[nostradamus-bot]

Nostradamus Risk Rating — Low

This PR is a performance-only refactor in the Go SDK WebSocket dispatch path (packages/sdk-go/websocket/client_dispatch.go), replacing a generic map-decode with a typed struct envelope; the change is functionally equivalent and no security-sensitive areas (auth, signing, fund-movement) are touched. The remaining diff is new benchmark tests and a minor regression test extension.