A previous PR made it so that the Go 1.17 collector locked only around
uses of rmSampleBuf, but really that means that Metric values may be
sent over the channel containing some values from future metrics.Read
calls. While generally-speaking this isn't a problem, we lose any
consistency guarantees provided by the runtime/metrics package.

Also, that optimization to not just lock around all of Collect was
premature. Truthfully, Collect is called relatively infrequently, and
its critical path is fairly fast (10s of µs). To prove it, this change
also adds a benchmark.

name            old time/op  new time/op  delta
GoCollector-16  43.7µs ± 2%  43.2µs ± 2%   ~     (p=0.190 n=9+9)

Note that because the benchmark is single-threaded it actually looks
like it might be getting *slightly* faster, because all those Collect
calls for the Metrics are direct calls instead of interface calls.

Signed-off-by: Michael Anthony Knyszek <[email protected]>

Replace `io.ReadAll` with `bytes.Buffer.ReadFrom`.
Both need to resize a buffer until they have finished reading;
the former increases by 1.25x each time while the latter uses 2x.

Also added a benchmark to demonstrate the benefit:
name             old time/op    new time/op    delta
Client/4KB-8       35.9µs ± 4%    35.3µs ± 3%     ~     (p=0.310 n=5+5)
Client/50KB-8      83.1µs ± 8%    69.5µs ± 1%  -16.37%  (p=0.008 n=5+5)
Client/1000KB-8     891µs ± 6%     750µs ± 0%  -15.83%  (p=0.016 n=5+4)
Client/2000KB-8    1.74ms ± 2%    1.35ms ± 1%  -22.72%  (p=0.008 n=5+5)

name             old alloc/op   new alloc/op   delta
Client/4KB-8       20.2kB ± 0%    20.4kB ± 0%   +1.26%  (p=0.008 n=5+5)
Client/50KB-8       218kB ± 0%     136kB ± 0%  -37.65%  (p=0.008 n=5+5)
Client/1000KB-8    5.88MB ± 0%    2.11MB ± 0%  -64.10%  (p=0.008 n=5+5)
Client/2000KB-8    11.7MB ± 0%     4.2MB ± 0%  -63.93%  (p=0.008 n=5+5)

name             old allocs/op  new allocs/op  delta
Client/4KB-8         75.0 ± 0%      72.0 ± 0%   -4.00%  (p=0.008 n=5+5)
Client/50KB-8         109 ± 0%        98 ± 0%  -10.09%  (p=0.079 n=4+5)
Client/1000KB-8       617 ± 0%       593 ± 0%   -3.89%  (p=0.008 n=5+5)
Client/2000KB-8     1.13k ± 0%     1.09k ± 0%   -3.27%  (p=0.008 n=5+5)

Signed-off-by: Bryan Boreham <[email protected]>

The Go runtime/metrics package currently exports extremely granular
histograms. Exponentially bucket any histogram with unit "seconds"
or "bytes" instead to dramatically reduce the number of buckets, and
thus the number of metrics.

This change also adds a test to check for expected cardinality to
prevent cardinality surprises in the future.

Signed-off-by: Michael Anthony Knyszek <[email protected]>

* Cut v1.12.1

Signed-off-by: Kemal Akkoyun <[email protected]>

* Apply review suggestions

Signed-off-by: Kemal Akkoyun <[email protected]>