adding ingestion delay metric

[Shvejan]

What this PR does:

Adds a new per-tenant native histogram metric cortex_ingester_ingestion_delay_seconds to track the
delay between when samples arrive at the ingester and their original timestamp. This metric helps
operators:

• Set appropriate rule_query_offset values based on actual ingestion delays
• Troubleshoot discrepancies between recording rules and underlying metrics
• Monitor and alert on ingestion lag per tenant

Which issue(s) this PR fixes:
Fixes #6748

Checklist

• Tests updated
• Documentation added
• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE],
  [ENHANCEMENT], [BUGFIX]
• docs/configuration/v1-guarantees.md updated if this PR introduces experimental flags

[friedrichg]

@Shvejan Let me know what you think

[friedrichg]

I think this line adds 15% overhead

go test -tags "netgo slicelabels" -run='^$' -bench=BenchmarkIngesterPush -benchtime=5s -count=3 ./pkg/ingester/...

it could be improved to 6/10% with something like:

  delayObserver := i.metrics.ingestionDelaySeconds.WithLabelValues(userID)                                                                                                                                                                      
...                                                                                                                                                                              
  observeDelay := func(timestampMs int64) {                                                                                                                                                                                                     
      if delayMs := time.Now().UnixMilli() - timestampMs; delayMs >= 0 {                                                                                                                                                                        
          delayObserver.Observe(float64(delayMs) / 1000.0)                                                                                                                                                                                      
      }                                                                                                                                                                                                                                         
  }  

The existing ingestedHistogramBuckets has the same per-sample WithLabelValues pattern and would benefit from the same treatment. But let's leave that for another PR