[bhupinder759]

Closes #123

Checklist for Pull Requests

• All tests pass (yarn test:all and yarn test:integration): unit tests and the non-DB integration tests pass locally; the backend and Postgres suites need Docker, which I don't have, so I'm relying on CI for those
• Code follows the style guide and passes lint checks
• Documentation is updated (README, docs, etc)
• Linked to corresponding issue, if applicable

Summary of Changes

Adds logger.adapter ("winston" | "pino") and logger.options, as discussed in #123:

await Sidequest.start({
  logger: { level: "info", adapter: "pino", options: { base: { service: "jobs" } } },
});

Design

• SidequestLogger (error/warn/info/debug(message, meta?)) is the strategy interface. createPinoLogger is the pino strategy, and loggerAdapters maps names to factories, so a new adapter is one factory plus one entry.
• Winston stays the internal logger. When an adapter is selected, its only transport forwards each entry (message, scope, metadata) to that strategy. logger(), the Logger type and all call sites are unchanged, and the default output is the same as before.
• The adapter is created in Engine.configure(), which runs in the app process, the forked engine process and each worker thread, so only serializable config crosses those boundaries. logger.options that can't be serialized (e.g. a timestamp function) fail at start instead of failing every job.
• pino is an optional peer dependency of @sidequest/core, imported only when selected, with an install hint if it's missing. Error values in metadata go through pino.stdSerializers.err.

Small behaviour changes

• The worker logs "Starting worker with provided configuration..." after engine.configure(), and validateConfig() runs after the logger is set up, so those lines also go through the selected adapter (and follow logger.level / logger.json).
• Engine.configure() registers the backend right after the config, before the new await, so a concurrent caller never sees the config without a backend.

Docs: installation ("Logger Adapters", with the warning that pino must be installed), logging guide ("Using pino"), configuration table.

How I tested

• yarn build, yarn lint, yarn format:check: pass.
• Unit tests per package: core 110 passed, engine 281 passed / 1 skipped, sidequest 62 passed.
• New tests/integration/pino-logger.integration.test.mjs (SQLite, no DB service): runs an app with adapter: "pino" in a child process and checks that every stdout line is pino JSON, including lines from the forked engine and a job's logger() call in a worker thread. It fails with the previous workers/main.ts order.
• Manually: default output unchanged; fork: false + runner: "inline" works.

[GiovaniGuizzo]

Thanks for the contribution, @bhupinder759! I want to check this manually too today when I have the time.