2.x: fix Single.timeout unnecessary dispose calls

[akarnokd]

This PR fixes the unnecessary dispose call towards the main source when the source terminates and reworks the internals to use less allocation and indirection.

[codecov]

Codecov Report

Merging #5586 into 2.x will increase coverage by 0.08%.
The diff coverage is 97.72%.

@@             Coverage Diff              @@
##                2.x    #5586      +/-   ##
============================================
+ Coverage     96.15%   96.23%   +0.08%     
- Complexity     5827     5828       +1     
============================================
  Files           632      632              
  Lines         41459    41467       +8     
  Branches       5742     5745       +3     
============================================
+ Hits          39863    39904      +41     
+ Misses          638      618      -20     
+ Partials        958      945      -13

Impacted Files	Coverage Δ	Complexity Δ
...tivex/internal/operators/single/SingleTimeout.java	98.27% <97.72%> (+2.27%)	2 <1> (ø)	⬇️
.../io/reactivex/internal/functions/ObjectHelper.java	94.73% <0%> (-5.27%)	20% <0%> (-1%)
...vex/internal/operators/single/SingleTakeUntil.java	91.8% <0%> (-3.28%)	2% <0%> (ø)
...tivex/internal/schedulers/TrampolineScheduler.java	92.3% <0%> (-2.57%)	6% <0%> (ø)
...activex/internal/observers/QueueDrainObserver.java	61.53% <0%> (-2.57%)	12% <0%> (-1%)
.../io/reactivex/disposables/CompositeDisposable.java	97.24% <0%> (-1.84%)	39% <0%> (-1%)
...java/io/reactivex/processors/PublishProcessor.java	98.29% <0%> (-1.71%)	45% <0%> (-1%)
...rnal/subscriptions/DeferredScalarSubscription.java	98.46% <0%> (-1.54%)	28% <0%> (-1%)
...vex/internal/operators/flowable/FlowableCache.java	92.61% <0%> (-1.35%)	7% <0%> (ø)
...perators/single/SingleFlatMapIterableFlowable.java	96.66% <0%> (-0.84%)	2% <0%> (ø)
... and 21 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c43229b...75b425d. Read the comment docs.

[artem-zinnatullin]

Mostly LGTM, but few comments

[artem-zinnatullin]

Shouldn't we deliver error to RxJavaPlugins.onError() in else block?

[akarnokd]

Indeed. Fixed.

[artem-zinnatullin]

Hmm, are you trying to avoid memory leak here? Reference to other is still held as a final field in SingleTimeout

I mean, final reference to other in TimeoutMainObserver will make code slightly more readable

[akarnokd]

The outer class has to reference other because of multiple subscribers. The inner observer has no use to the other once it has switched to it. There was an issue that some time ago that other was some cached resource that didn't go away when the user wanted it. This will somewhat help in that situation.

[artem-zinnatullin]

nit: I know you often use arrays for that, but AtomicInteger kinda much more readable when you need a reference to single "mutable" int :)