Can we add configuration on dropping raw vectors from quantized formats after some period of time?

### Description

Tangentially related to: https://github.com/apache/lucene/issues/13158

But, I have observed, that as the corpus reaches a fairly large size, the actual quantiles aren't changing much at all during segment merges. This is tricky to fully measure and make a promise about lossiness (users can always just start throwing garbage that shakes up the whole world). But if the data isn't a "bad actor", quantiles and quantization buckets become fairly stable over time.

Maybe we should add a configuration option, or a new codec, or some way to drop the raw floating point vectors. The 4x reduction in disk usage would be really nice for many use-cases.

I am not 100% sure how this would look (a threshold provided by the user, or we just do it based on internal statistics).

The issue names no files, tests, or entry points. First investigate quantized vector handling during segment merges and compare the configuration, codec, and internal-statistics alternatives. Done should be defined as a clear way to drop raw floating-point vectors after a stability threshold while documenting the resulting lossiness.

Written by the indexing model from the issue text.