Detach ingester series from gRPC buffers to reduce heap

[justinjung04]

What this PR does:
Detach series label and chunk data from gRPC unmarshal buffers in the ingester streaming path (distributor_queryable.go), allowing the Go GC to reclaim receive buffers to reduce heap usage. This is the ingester-side counterpart to #7519, which applied the same fix for the store-gateway streaming path.

Specifically:

• Use FromLabelAdaptersToLabelsWithCopy instead of FromLabelAdaptersToLabels (unsafe cast that pins the buffer)
• Copy chunk Data byte slices into independent allocations via detachChunksFromBuffer

Without this change, long-running queries hold references into gRPC receive buffers through label strings and chunk data, preventing GC from reclaiming them and causing heap to grow under heavy ingester query load.

Benchmark (100 series, 1KB chunks each):

	ns/op	B/op	allocs/op
current (no detach)	8,000	23,416	402
this PR	31,500	161,016	902

The per-call cost of copying is ~137KB additional short-lived allocations per 100 series. The tradeoff is that gRPC receive buffers (~32KB+ each) are freed immediately after processing instead of being pinned for the entire query lifetime. Under concurrent query load, this prevents heap accumulation from unreclaimable buffers.

Which issue(s) this PR fixes:
N/A

Checklist

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