feat(build): feature-sided builds — server/client sides independently selectable (ADR-039 Amendment 1)

Split the feature graph so consumers pulling only the import adapters
(from_openapi / from_jsonschema / from_mcp) no longer compile the axum
/ hyper server stack, and server-only deployments no longer compile
reqwest. One crate, one import path — sides cut by features, not by a
crate split.

Feature graph:
- server (default): axum host, gateway, WS upgrade, to_openapi, to_mcp
- client (default): client host, forward, from_jsonschema, from_openapi
- openapi: shared OpenAPISpec model (implied by both sides)
- mcp: from_mcp needs client, to_mcp needs server
- wss: tungstenite transport (from_wss); tungstenite half of the
  shared WS↔byte-stream adapter
- h2/http1: hyper protocol features; imply server

Wire-contract neutral: gateway endpoints, ALPNs, and all public API
shapes unchanged; defaults keep both sides on.

Supporting changes:
- forward.rs drops its axum::body::Bytes type leak (bytes crate types)
- bounded_join + error-echo caps move to input_validation (usable by
  both sides; openapi_spec no longer imports from forward)
- byte_adapter: axum flavor compiles under server, tungstenite under
  wss; the generic pumps stay shared (WS-11)
- input_validation / openapi_spec import-only internals gated to the
  side that consumes them
- http-body-util moves to dev-dependencies (was test-only)
- integration-test required-features updated for the new sides
- from_wss unit tests (axum producer harness) gated to server

Verified: cargo test (defaults, 453) and --all-features (575) pass;
lean side builds (client / server / client,mcp / client,wss /
server,wss / openapi-only) build clean with zero warnings;
clippy -D warnings clean across all feature combinations; fmt clean.

docs/architecture/decisions/039-http-server-and-client-host-colocated.md

@@ -138,6 +138,30 @@ concerns that make splitting them counterproductive:
    argument is weaker — but the current design (ADR-036, ADR-023)
    has them sharing the mapping.
 
+## Amendment 1: feature-sided builds (`server` / `client`)
+
+The colocation decision stands — one crate, one import path — but the
+"the compile cost is paid once per workspace" mitigation proved
+insufficient: a consumer that only imports (consumer side) was forced
+to compile the axum/hyper server stack, and vice versa. The feature
+graph now sides the crate without splitting it:
+
+- `server` (default) — the axum `Router` host, the gateway routes,
+  the WS upgrade path, `to_openapi`, `to_mcp` (with `mcp`).
+- `client` (default) — the outbound client host, the `from_*` import
+  adapters (`from_wss` additionally needs `wss`).
+- `openapi` — the shared `OpenAPISpec` document model (implied by
+  both sides; `from_openapi` needs it, `to_openapi` needs it).
+- `mcp` — `from_mcp` requires `client`, `to_mcp` requires `server`.
+- `wss` — the tungstenite WS transport (`from_wss`; the tungstenite
+  flavor of the shared WS↔byte-stream adapter).
+- `h2` / `http1` — hyper protocol features; imply `server`.
+
+Default features keep both sides on (zero behavior change); lean
+builds use `default-features = false` + the side they need. The
+wire contract (gateway endpoints, ALPNs) is untouched — this is a
+dependency-graph change, not a surface change.
+
 ## References
 
 - [ADR-003](003-crate-decomposition.md) — crate decomposition (this