Reduce memory allocations in bucketBlock.readChunkRange()

[pracucci]

• I added CHANGELOG entry for this change.
• Change is not relevant to the end user.

Changes

We're running the Cortex blocks storage at an increasingly large scale. We're currently investigating how to reduce the memory utilisation in the store-gateway and a low hanging fruit I've noticed is that bucketBlock.readChunkRange() waste memory allocations to grow the buffer if the requested length is larger than the largest bucket in the chunk pool (which is 32MB):

The reason is that to ensure that io.Copy() doesn't grow the buffer, the buffer needs to be at least bytes.MinRead larger than the actual size to write.

This PR fixes it.

I've also added a benchmark, comparing it with master:

benchmark                                  old ns/op     new ns/op     delta
BenchmarkBucketBlock_readChunkRange-12     89789         70651         -21.31%

benchmark                                  old allocs     new allocs     delta
BenchmarkBucketBlock_readChunkRange-12     12             12             +0.00%

benchmark                                  old bytes     new bytes     delta
BenchmarkBucketBlock_readChunkRange-12     296676        204791        -30.97%

Verification

Bencharmk.

[GiedriusS]

Great work!