ref: centralize frames tracker access in SentryDependencyContainer

[armcknight]

In service of reworking SentryProfiler's management of GPU frame data in #3090, I need to simplify how it keeps a reference to the frames tracker.

This has bugged me for some time because we vend a singleton of the frames tracker, but then various other objects keep and pass references to it, which is unnecessary; it can just be accessed via SentryDependencyContainer. This seemed like a good opportunity to fix it everywhere.

#skip-changelog

[codecov]

Codecov Report

Merging #3119 (cff6328) into main (6bc5ae5) will decrease coverage by 0.016%.
The diff coverage is 100.000%.

Additional details and impacted files

@@              Coverage Diff              @@
##              main     #3119       +/-   ##
=============================================
- Coverage   89.040%   89.025%   -0.016%     
=============================================
  Files          500       500               
  Lines        53910     53898       -12     
  Branches     19288     19278       -10     
=============================================
- Hits         48002     47983       -19     
- Misses        4951      4953        +2     
- Partials       957       962        +5     

Impacted Files	Coverage Δ
Sources/Sentry/SentryFramesTracker.m	98.449% <ø> (-0.102%)	⬇️
.../Sentry/SentryUIViewControllerPerformanceTracker.m	98.800% <ø> (-0.005%)	⬇️
Sources/Sentry/PrivateSentrySDKOnly.m	88.888% <100.000%> (ø)
Sources/Sentry/SentryDependencyContainer.m	93.478% <100.000%> (+0.414%)	⬆️
Sources/Sentry/SentryFramesTrackingIntegration.m	100.000% <100.000%> (ø)
Sources/Sentry/SentryProfiler.mm	85.350% <100.000%> (-0.245%)	⬇️
Sources/Sentry/SentryTimeToDisplayTracker.m	100.000% <100.000%> (ø)
Sources/Sentry/SentryTracer.m	96.268% <100.000%> (-0.747%)	⬇️
...SentryProfilerTests/SentryProfilerSwiftTests.swift	97.033% <100.000%> (ø)
...racking/SentryFramesTrackingIntegrationTests.swift	98.630% <100.000%> (ø)
... and 5 more

... and 12 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 6bc5ae5...cff6328. Read the comment docs.

[github-actions]

Performance metrics 🚀

	Plain	With Sentry	Diff
Startup time	1244.39 ms	1256.74 ms	12.35 ms
Size	20.76 KiB	401.56 KiB	380.80 KiB

Baseline results on branch: main

Startup times

Revision	Plain	With Sentry	Diff
6943de0	1230.02 ms	1235.32 ms	5.30 ms
20163bb	1248.92 ms	1258.48 ms	9.56 ms
25a5e8b	1249.18 ms	1268.42 ms	19.24 ms
8526e93	1228.24 ms	1239.14 ms	10.90 ms
154f795	1250.38 ms	1274.54 ms	24.16 ms
45d3ca5	1238.53 ms	1263.09 ms	24.55 ms
e1eed6b	1259.08 ms	1270.57 ms	11.49 ms
7e8d5fd	6249.23 ms	6253.90 ms	4.67 ms
dcec216	1238.94 ms	1261.06 ms	22.12 ms
a2af9fa	1236.62 ms	1253.12 ms	16.50 ms

App size

Revision	Plain	With Sentry	Diff
6943de0	20.76 KiB	393.33 KiB	372.57 KiB
20163bb	20.76 KiB	426.82 KiB	406.06 KiB
25a5e8b	20.76 KiB	436.33 KiB	415.57 KiB
8526e93	20.76 KiB	420.22 KiB	399.47 KiB
154f795	20.76 KiB	435.25 KiB	414.49 KiB
45d3ca5	20.76 KiB	427.54 KiB	406.78 KiB
e1eed6b	20.76 KiB	432.17 KiB	411.41 KiB
7e8d5fd	20.76 KiB	435.50 KiB	414.74 KiB
dcec216	20.76 KiB	432.88 KiB	412.11 KiB
a2af9fa	20.76 KiB	432.87 KiB	412.11 KiB

Previous results on branch: armcknight/ref/framestracker

Startup times

Revision	Plain	With Sentry	Diff
c15ad1e	1212.96 ms	1230.06 ms	17.10 ms

App size

Revision	Plain	With Sentry	Diff
c15ad1e	20.76 KiB	401.56 KiB	380.80 KiB

[armcknight]

It turns out that with a constant instance property, it was initialized as part of a pre-test process, and then when the tests ran, the dependency container shared instance was constructed again and had a difference instance of the frames tracker, so the tests would fail. Making it lazy defers initialization until the test process itself is running.

[brustolin]

LGTM