Reuse profile rewriter during ingestion

When ingesting pprof profiles, we need to remap most IDs to already ingested symbols.

Right now the code creates a new set of rewriter maps for each symbols types. (strings, functions, locations, mapping, stacktraces)

https://github.com/grafana/phlare/blob/7d20115c15e650afc60bc4c0334b403e2550e434/pkg/phlaredb/deduplicating_slice.go#L177

Those are then added kept for the scope of the request using:

https://github.com/grafana/phlare/blob/7d20115c15e650afc60bc4c0334b403e2550e434/pkg/phlaredb/deduplicating_slice.go#LL228C15-L228C15

To rewrite further symbols ids encountered.

https://github.com/grafana/phlare/blob/7d20115c15e650afc60bc4c0334b403e2550e434/pkg/phlaredb/deduplicating_slice.go#LL182C2-L186C3

Because we expect at some point to rewrite everything we should probably use arrays of the size of the input slice that can be re-used via a `sync.Pool`. Ultimately I think we should pool and reuse rewriters. We should have a different pool for each type though since the size of each types varies a lot.

For example we expect thousands of stracktraces, but only few mappings.