Introduce max bucket count limit for native histograms

[yeya24]

Part of #6052

Is your feature request related to a problem? Please describe.

At ingestion path, the native histograms data ingested might have very high resolution and contains a large amount of buckets. Ingesting such kind of histogram may easily overload Cortex Ingester and cause potential outage.

Describe the solution you'd like

We can follow the same behavior as what Prometheus does during scrape time.

Introduce a native_histogram_bucket_limit config to be set at Distributor/Ingester. If the input histogram exceeds this limit, we can keep reducing its resolution until bucket count within the limit or invalid schema.