Add internal EXPLAIN ANALYZE metric level

### Is your feature request related to a problem or challenge?

Some execution metrics are useful for low-level operator debugging but are too noisy or high-cardinality for the default `EXPLAIN ANALYZE` output. Today the `summary` and `dev` levels do not provide a clean place for these internal development metrics.

### Describe the solution you would like

Add an `internal` `datafusion.explain.analyze_level` that includes `summary`, `dev`, and low-level internal metrics. Operators can use this level for targeted debugging metrics while keeping the default `dev` output concise and avoiding unnecessary metric registration overhead.

### Describe alternatives you have considered

Add operator-specific flags or environment variables for granular metrics. That makes metrics harder to discover and does not compose well across operators.

### Additional context

`RepartitionExec` granular timings are the first intended use case for this level.

Start by tracing how datafusion.explain.analyze_level is parsed and how EXPLAIN ANALYZE selects metrics, then inspect RepartitionExec as the stated first use case. Done means an internal level includes summary, dev, and low-level internal metrics while dev remains concise and avoids unnecessary metric registration overhead.

Written by the indexing model from the issue text.