feat(profiling): starting and stopping continuous profiler

[armcknight]

for #3555; #skip-changelog

After some more discussion, we're only going to offer manually starting and stopping the profiler (besides app launch profiling starting it automatically before any app code can actually run; but that must at least be manually configured by customers by setting the SentryOption for it).

We originally planned to automatically start/stop with session start/stop that respected the profiles sampling definitions, but we won't move forward with that right now. Manual start/stop will not respect profiling sample rates, calls will always start the profiler if one isn't already running.

Closing the SDK will stop any profiler that's running.

This cleans up anything that tied profiles to sessions, and adds more tests around various start/stop scenarios.

[codecov]

Codecov Report

Attention: Patch coverage is 97.50000% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 90.891%. Comparing base (c838244) to head (d0c8e78).
Report is 1 commits behind head on main.

Additional details and impacted files

@@              Coverage Diff              @@
##              main     #3937       +/-   ##
=============================================
+ Coverage   90.858%   90.891%   +0.032%     
=============================================
  Files          594       594               
  Lines        46328     46351       +23     
  Branches     16599     16610       +11     
=============================================
+ Hits         42093     42129       +36     
+ Misses        4165      4152       -13     
  Partials        70        70               

Files	Coverage Δ
SentryTestUtils/ClearTestState.swift	100.000% <100.000%> (ø)
Sources/Sentry/Profiling/SentryLaunchProfiling.m	61.842% <100.000%> (ø)
Sources/Sentry/SentryHub.m	98.554% <100.000%> (ø)
Sources/Sentry/SentrySDK.m	86.585% <100.000%> (+0.249%)	⬆️
Sources/Sentry/SentrySampling.m	92.682% <ø> (ø)
Sources/Sentry/SentryTracer.m	96.897% <100.000%> (+0.007%)	⬆️
...yProfilerTests/SentryContinuousProfilerTests.swift	100.000% <100.000%> (ø)
...SentryProfilerTests/SentryProfileTestFixture.swift	99.401% <100.000%> (+0.003%)	⬆️
Sources/Sentry/SentryProfiler.mm	91.025% <92.857%> (+0.116%)	⬆️

... and 8 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 c838244...d0c8e78. Read the comment docs.

[philipphofmann]

LGTM

[github-actions]

Performance metrics 🚀

	Plain	With Sentry	Diff
Startup time	1233.61 ms	1254.11 ms	20.50 ms
Size	21.58 KiB	629.82 KiB	608.24 KiB

Baseline results on branch: main

Startup times

Revision	Plain	With Sentry	Diff
deeb22c	1233.90 ms	1250.19 ms	16.29 ms
3a6495e	1227.61 ms	1239.22 ms	11.60 ms
742d4b6	1196.56 ms	1216.54 ms	19.98 ms
89491ad	1222.12 ms	1231.96 ms	9.83 ms
1a4da1b	1222.14 ms	1239.50 ms	17.36 ms
236d8a8	1232.33 ms	1255.55 ms	23.22 ms
983de17	1260.57 ms	1263.68 ms	3.11 ms
1b09e3f	1227.24 ms	1242.19 ms	14.95 ms
3cba0e8	1231.20 ms	1251.50 ms	20.30 ms
59c1b97	1230.70 ms	1257.12 ms	26.42 ms

App size

Revision	Plain	With Sentry	Diff
deeb22c	21.58 KiB	612.11 KiB	590.53 KiB
3a6495e	21.58 KiB	422.66 KiB	401.08 KiB
742d4b6	21.58 KiB	546.20 KiB	524.62 KiB
89491ad	21.58 KiB	417.89 KiB	396.30 KiB
1a4da1b	21.58 KiB	418.33 KiB	396.75 KiB
236d8a8	21.58 KiB	418.70 KiB	397.12 KiB
983de17	22.84 KiB	403.19 KiB	380.34 KiB
1b09e3f	21.58 KiB	614.93 KiB	593.34 KiB
3cba0e8	22.84 KiB	403.18 KiB	380.34 KiB
59c1b97	21.58 KiB	547.63 KiB	526.05 KiB