fix: reserve chat hook dispatch capacity for running turns (#27656)

## Context

Follow-up fix from live UAT of the merged chat lifecycle hooks stack
(#27430). Its companion UAT fix (#27655) has merged, so this targets
`main` directly.

## Why?

UAT measured a burst of 1,500 concurrent chat creations against a
consumer with 1.2s latency. 255 were admitted and 1,245 got `502
hook_dispatch_failed (over_capacity)`, which is correct fail-closed
behavior. The collateral wasn't: the same burst failed 24 `stop`
dispatches, parking chats that had already been admitted and had already
executed tools. One 256-slot semaphore served every event, so new-work
admission could take every slot and kill turns in flight.

Callers now classify each dispatch as admission or generation, and
admission draws from a 192-slot gate held *before* the shared pool. At
least 64 shared slots stay reachable only by dispatches for work a chat
already admitted. The dispatcher is per `coderd` replica, so these
limits are per replica, not deployment-wide, and the docs say so.

**The caller classifies, not the event type.** Event type isn't a
reliable proxy in either direction: a subagent spawn dispatches
`user_prompt_submit` from inside a running turn, and the edit path
dispatches `session_start` at admission time. `CapacityClassUnset` is
rejected in `Dispatch`, so a new call site fails closed rather than
silently inheriting a share.

**Acquisition order is load-bearing.** Admission takes its own gate
first. Taking a shared slot first would let admissions queued on the
gate occupy the very capacity the reserve protects. `acquireCapacity` is
the only path that takes either pool, so the order can't be bypassed.

## What this does not guarantee

Nothing bounds how many turns generate concurrently, so the 192/64 split
is a judgement call, not a derived ceiling. This stops an *admission*
burst from consuming every slot; it does not make the remainder
sufficient. A large enough generation load can still exhaust the reserve
and error a running chat. The docs say so explicitly rather than
promising a guarantee the code doesn't deliver.

Generation can now take all 256 slots, so generation traffic starves
admission harder than before. That's the intended priority: rejecting a
new prompt is recoverable, ending a turn that already ran tools is not.

## Testing

Red-green proved both new tests. Removing the release-on-failure path
fails `RefusedSharedAcquireReleasesAdmission` deterministically;
removing the expired-deadline check fails
`ExpiredDeadlineRefusesFreeSlot` in 18/30 runs.

That deadline check fixes a real race found in review. `acquire`
previously shared one `time.Timer` across both acquires. Because
`select` picks a ready case at random, an admission dispatch could take
a slot after its capacity deadline had passed. Measured over 300 trials:
135 late acquisitions, worst overshoot 2.1ms. `acquire` now takes an
absolute deadline and refuses an expired one before selecting, which
measures 0/300.

Go: `coderd/x/agenthooks/...` and `coderd/x/chatd/...`, plus `-race
-count=3` on the dispatcher.

> Mux opened this PR on Mike's behalf.

coderd/x/chatd/ARCHITECTURE.md

@@ -915,6 +915,8 @@ The consumer can observe activity, add model-only or user-visible context, repla
 
 Lifecycle hooks fail closed. If the consumer cannot be reached or returns an invalid response, Coder stops the triggering operation rather than continuing without the consumer's decision. Affected chats can enter an error state until the consumer recovers or hooks are disabled.
 
+Concurrent dispatches are capped per replica, and each dispatch declares whether it admits new work into a chat or belongs to work a chat already admitted. Admission can hold only part of the cap, so a burst of new submissions cannot consume the capacity that already-admitted work depends on. The caller declares this, because the event type does not determine it: a subagent spawn submits a prompt from inside a running turn, and editing a message starts a session at admission time.
+
 Coder stores no hook-specific dispatch or decision state. Delivery is best-effort and can duplicate, and a failed dispatch is never redelivered, so the consumer owns durable policy state, audit records, and deduplication based on stable event identifiers.
 
 # Stream loop