Optimize fixedsize reservoir

[dashpole]

Depends on #7441, #7443

This improves the concurrent performance of the fixed size reservoir's Offer function by 4x (i.e. 75% reduction). This improves the performance of Measure() for fixed-size reservoirs by 60% overall.

Accomplish this by:

• using a single atomic for count and next. This assumes that both can fit in a uint32.
• only use a lock to guard changing w and next together.

Offer benchmarks:

                           │   main.txt   │           fixedsize.txt            │
                           │    sec/op    │   sec/op     vs base               │
FixedSizeReservoirOffer-24   185.25n ± 4%   45.58n ± 1%  -75.40% (p=0.002 n=6)

Measure benchmarks:

                                                                          │   main.txt   │            fixedsize.txt            │
                                                                          │    sec/op    │    sec/op     vs base               │
SyncMeasure/NoView/ExemplarsEnabled/Int64Counter/Attributes/0-24            175.45n ± 6%   67.01n ±  9%  -61.81% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Int64Counter/Attributes/1-24            170.25n ± 1%   69.82n ±  6%  -58.99% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Int64Counter/Attributes/10-24           167.40n ± 2%   64.52n ± 10%  -61.46% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Float64Counter/Attributes/0-24          173.55n ± 0%   69.17n ± 12%  -60.14% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Float64Counter/Attributes/1-24          169.50n ± 1%   68.55n ±  5%  -59.56% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Float64Counter/Attributes/10-24         166.95n ± 1%   65.82n ±  6%  -60.58% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Int64UpDownCounter/Attributes/0-24      168.85n ± 1%   67.99n ± 11%  -59.73% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Int64UpDownCounter/Attributes/1-24      173.50n ± 1%   66.69n ±  2%  -61.56% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Int64UpDownCounter/Attributes/10-24     171.30n ± 5%   67.73n ±  8%  -60.46% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Float64UpDownCounter/Attributes/0-24    168.90n ± 2%   67.69n ±  9%  -59.92% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Float64UpDownCounter/Attributes/1-24    173.35n ± 2%   68.25n ±  9%  -60.63% (p=0.002 n=6)
SyncMeasure/NoView/ExemplarsEnabled/Float64UpDownCounter/Attributes/10-24   172.95n ± 2%   70.90n ±  7%  -59.01% (p=0.002 n=6)
geomean                                                                      171.0n        67.83n        -60.33%

[codecov]

Codecov Report

❌ Patch coverage is 90.24390% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 86.2%. Comparing base (87c8b6a) to head (d6a538f).

Files with missing lines	Patch %	Lines
sdk/metric/exemplar/fixed_size_reservoir.go	90.2%	2 Missing and 2 partials ⚠️

Additional details and impacted files

@@          Coverage Diff          @@
##            main   #7447   +/-   ##
=====================================
  Coverage   86.2%   86.2%           
=====================================
  Files        302     302           
  Lines      21991   22002   +11     
=====================================
+ Hits       18968   18980   +12     
+ Misses      2642    2641    -1     
  Partials     381     381           

Files with missing lines	Coverage Δ
sdk/metric/exemplar/fixed_size_reservoir.go	91.8% <90.2%> (-3.0%)	⬇️

... and 3 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[dashpole]

@MrAlias @dmathieu this is ready for review!

[MrAlias]

Should the lock be held here? Why not within advance? This seems very prone to deadlocks.

[dashpole]

We don't hold the lock within advance because we also call advance() from within reset(), which needs to hold the lock for all operations.

It shouldn't be prone to deadlocks because both locations where we acquire the lock (reset() and Offer()) don't block on anything else while they hold the lock.

[MrAlias]

Missing test cases for when k is 0 and greater than math.Maxuint32.

[dashpole]

Added a test for k < 0. I tried to add a test case for greater than maxuint32, but I can't allocate a slice that large on my computer. I could use a bunch of mocks and such to test the if statement, but i'm pretty sure it isn't worth it.