feat(node): update log level via admin API

[spalladino]

Allow changing the log level of a running node at runtime through the node admin API, instead of only via the LOG_LEVEL env var at startup.

Context

Diagnosing issues on a live node often requires more verbose logging, but until now the only way to change the log level was restarting the node with a different LOG_LEVEL — losing the state you were trying to observe.

Approach

• Adds an optional logLevel key to AztecNodeAdminConfig, so it flows through the existing aztecAdmin_setConfig / aztecAdmin_getConfig methods with their auth and merge semantics. The value uses the same spec format as the LOG_LEVEL env var (e.g. debug;trace:sequencer,p2p) and is validated at the RPC boundary.
• Pino child loggers bake in their level at creation and do not follow later changes to the parent, so @aztec/foundation/log now keeps a registry of all live child loggers (weakly held via WeakRef/FinalizationRegistry, since some paths create short-lived per-tx loggers). A new setLogLevel re-levels the root and every registered child per the new filters; loggers created afterwards pick up the new levels too. libp2p component loggers go through the same registry.
• No transport changes needed: all transports are already registered at trace, so gating happens entirely at the logger.

The change only affects the node's own process — spawned prover agents and external binaries keep their configuration — and is not persisted: a restart reverts to the LOG_LEVEL env var.

API changes

AztecNodeAdmin.setConfig accepts an optional logLevel string and getConfig reports the current spec. @aztec/foundation/log exports new setLogLevel, getLogLevel, formatLogLevelSpec, and isValidLogLevelSpec functions.

Example:

curl -s http://node:8880/ -H 'x-api-key: ...' \
  -d '{"jsonrpc":"2.0","id":1,"method":"aztecAdmin_setConfig","params":[{"logLevel":"debug;trace:p2p"}]}'