fix: cpu usage readings for profiling

[armcknight]

We noticed as we were iterating that we weren't calculating the CPU usage correctly. After fixing it, we realized that this reading doesn't report CPU usage per core only for the process using the API; it shows usage across all processes. We decided to use the API that reports a total CPU usage per thread, to guarantee process confinement, and sum these into a single number. This has a benefit of matching how Android reports this metric.

💚 How did you test it?

Visual inspection of the Xcode debugger and matching it to what we see in the flamechart viewer.

Xcode debugger:

From the profile:

A bit hard to compare due to limitations sizing them to match scale, but the profiler's version basically matches the same trends, except for the one spike shown in the Xcode debugger. Also watched the Xcode "speedometer" with the single CPU % and numbers logged from the profiling code throughout the test, and they basically matched while I changed the amount of CPU work to do in the test app:

[codecov]

Codecov Report

Merging #3214 (6e9ef71) into main (b066509) will decrease coverage by 0.018%.
The diff coverage is 69.811%.

Additional details and impacted files

@@              Coverage Diff              @@
##              main     #3214       +/-   ##
=============================================
- Coverage   89.199%   89.182%   -0.018%     
=============================================
  Files          502       502               
  Lines        54081     54041       -40     
  Branches     19413     19394       -19     
=============================================
- Hits         48240     48195       -45     
- Misses        4988      4991        +3     
- Partials       853       855        +2     

Files Changed	Coverage Δ
...SentryProfilerTests/SentrySystemWrapperTests.swift	85.714% <0.000%> (+2.380%)	⬆️
Sources/Sentry/SentrySystemWrapper.mm	60.655% <64.705%> (-13.258%)	⬇️
Sources/Sentry/SentryMetricProfiler.mm	92.380% <85.714%> (-2.578%)	⬇️
...SentryProfilerTests/SentryProfilerSwiftTests.swift	97.517% <100.000%> (-0.014%)	⬇️

... and 16 files with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b066509...6e9ef71. Read the comment docs.

[github-actions]

Performance metrics 🚀

	Plain	With Sentry	Diff
Startup time	1242.27 ms	1254.71 ms	12.44 ms
Size	22.84 KiB	402.63 KiB	379.79 KiB

Baseline results on branch: main

Startup times

Revision	Plain	With Sentry	Diff
b385962	1195.85 ms	1221.63 ms	25.78 ms
1bf8571	1215.31 ms	1232.48 ms	17.17 ms
60dd0f5	1238.98 ms	1254.48 ms	15.50 ms
6ad07ae	1240.76 ms	1242.98 ms	2.22 ms
c319795	1205.12 ms	1231.20 ms	26.08 ms
8f397a7	1251.82 ms	1268.34 ms	16.52 ms
46f5eb8	1212.27 ms	1231.42 ms	19.15 ms
dc0db9e	1246.06 ms	1260.46 ms	14.40 ms
25e65a5	1249.08 ms	1277.78 ms	28.70 ms
b8dd0fc	1235.57 ms	1253.12 ms	17.55 ms

App size

Revision	Plain	With Sentry	Diff
b385962	20.76 KiB	399.69 KiB	378.93 KiB
1bf8571	20.76 KiB	437.12 KiB	416.36 KiB
60dd0f5	20.76 KiB	393.37 KiB	372.61 KiB
6ad07ae	20.76 KiB	424.69 KiB	403.93 KiB
c319795	20.76 KiB	431.99 KiB	411.22 KiB
8f397a7	20.76 KiB	420.55 KiB	399.79 KiB
46f5eb8	20.76 KiB	432.37 KiB	411.61 KiB
dc0db9e	20.76 KiB	419.62 KiB	398.86 KiB
25e65a5	22.85 KiB	403.19 KiB	380.34 KiB
b8dd0fc	20.76 KiB	401.39 KiB	380.63 KiB

Previous results on branch: armcknight/fix/cpu-usage-readings

Startup times

Revision	Plain	With Sentry	Diff
9ce25d7	1239.69 ms	1251.94 ms	12.25 ms
ed8a1f8	1240.08 ms	1261.12 ms	21.04 ms

App size

Revision	Plain	With Sentry	Diff
9ce25d7	22.84 KiB	402.82 KiB	379.97 KiB
ed8a1f8	22.84 KiB	402.65 KiB	379.80 KiB

[brustolin]

LGTM