store: fix chunk pool memory leak

[abursavich]

I experienced a slow drip of "key does not exist" errors, presumably due to compactions, then nearly all get_range requests began to fail with "pool exhausted" errors.

caller	err
bucket.go:871	rpc error: code = Aborted desc = fetch series for block 01DM8ACH11QQQESTY0VET5PXKS: preload chunks: read range for 1: get range reader: The specified key does not exist.
bucket.go:871	rpc error: code = Aborted desc = fetch series for block 01DMA1C7D151KM6H1DGY498WWQ: preload chunks: read range for 6: allocate chunk bytes: pool exhausted

Changes

Return the byte slice to the pool on error paths.

Verification

N/A

[GiedriusS]

Thanks for catching this, we are definitely forgetting to put the contents back into the pool here! All of the other places seem OK to me.

[bwplotka]

Nice catch!