docs(worker): correct tier from 2 to 3 and fix dependency tier refs

worker depends on launcher (now correctly Tier 2) and logz (Tier 1),
placing it at Tier 3. The previous docs cited launcher as Tier 1 and
logz as Tier 0, both of which were wrong.

CHANGELOG.md

@@ -0,0 +1,29 @@
+# Changelog
+
+All notable changes to this module will be documented in this file.
+
+The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
+and this module adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
+
+## [0.9.0] - 2026-03-18
+
+### Added
+
+- `Task` type — `func(ctx context.Context) error`; the unit of work dispatched to the pool
+- `Config` struct — pool settings loaded from environment variables: `WORKER_POOL_SIZE` (number of concurrent goroutines, default `5`), `WORKER_BUFFER_SIZE` (task queue channel capacity, default `100`), `WORKER_TASK_TIMEOUT` (per-task context deadline; `0` means no deadline, default `0s`), `WORKER_SHUTDOWN_TIMEOUT` (time to wait for workers to drain on stop, default `30s`)
+- `Provider` interface — `Dispatch(task Task) bool`; for callers that only dispatch tasks; returns `false` immediately when the queue is full (backpressure, non-blocking)
+- `Component` interface — embeds `launcher.Component` and `Provider`; the full lifecycle-managed surface registered with the launcher
+- `New(logger logz.Logger, cfg Config) Component` — constructor; applies safe defaults (`PoolSize <= 0` → 5, `BufferSize <= 0` → 100); returns a `Component` ready for `lc.Append`
+- `OnInit` — logs pool configuration; initialises the buffered task channel
+- `OnStart` — spawns `PoolSize` worker goroutines, each ranging over the task channel
+- `OnStop` — closes the task channel (drain signal), cancels the pool context, then waits up to `ShutdownTimeout` for all goroutines to finish; logs an error on timeout but returns `nil` so the launcher continues
+- Per-task timeout — when `TaskTimeout > 0`, each worker creates a `context.WithTimeout` child before invoking the task; tasks also receive a cancellation signal when the pool is stopping via the pool context
+- Error logging — task errors are logged with the worker ID; errors are not surfaced to the dispatcher
+
+### Design Notes
+
+- A single buffered `chan Task` is shared by all workers; closing it during `OnStop` is the drain signal, avoiding a separate done channel or additional synchronisation primitives.
+- `Dispatch` is deliberately non-blocking: a `false` return means the task has been dropped, not queued; the caller owns the retry or overflow decision, keeping backpressure handling out of the pool itself.
+- `Provider` / `Component` split follows the framework pattern: inject `Provider` into callers that only dispatch tasks, inject `Component` only at the lifecycle registration site.
+
+[0.9.0]: https://code.nochebuena.dev/go/worker/releases/tag/v0.9.0