runtime: allow experimental trace batches to be reused
Currently, we can only cache regular trace event buffers on each M. As
a result, calling unsafeTraceExpWriter will, in effect, always return
a new trace batch, with all of the overhead that entails.
This extends that cache to support buffers for experimental trace
data. This way, unsafeTraceExpWriter can return a partially used
buffer, which the caller can continue to extend. This gives the caller
control over when these buffers get flushed and reuses all of the
existing trace buffering mechanism.
This also has the consequence of simplifying the experimental batch
infrastructure a bit. Now, traceWriter needs to know the experiment ID
anyway, which means there's no need for a separate traceExpWriter
type.
Change-Id: Idc2100176c5d02e0fbb229dc8aa4aea2b1cf5231
Reviewed-on: https://go-review.googlesource.com/c/go/+/594595
Auto-Submit: Austin Clements <austin@google.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>