[PLAT-11315] Discard unfinished spans when the app goes into the background

[kstenerud]

Goal

Spans that are still open when the app is backgrounded will end up with ridiculously long durations. Throw out unended spans when backgrounding the app instead.

Testing

Since we can't automate app backgrounding in a reliable way, I tested this manually.

The new weak span list component also has unit tests.

[github-actions]

BugsnagPerformance.framework binary size increased by 22,696 bytes from 466,968 to 489,664

    FILE SIZE        VM SIZE    
 --------------  -------------- 
 +40e2% +15.5Ki +56e2% +15.5Ki    [__DATA]
  +2.4% +3.82Ki  +2.4% +3.82Ki    String Table
  +2.7% +2.25Ki  +2.7% +2.25Ki    Symbol Table
  +2.2% +2.09Ki  +2.2% +2.09Ki    __TEXT,__text
  +3.9%    +416  +3.9%    +416    __DATA,__objc_const
  +3.0%    +276  +3.0%    +276    __TEXT,__gcc_except_tab
  +1.7%    +176  +1.7%    +176    __TEXT,__objc_methname
  +1.5%    +128  +1.5%    +128    __TEXT,__objc_stubs
  +3.1%    +120  +3.1%    +120    __TEXT,__unwind_info
  +0.9%     +80  +0.9%     +80    [9 Others]
  +7.1%     +80  +7.1%     +80    __DATA,__objc_data
  +1.8%     +80  +1.8%     +80    __DATA_CONST,__const
  +2.8%     +80  +2.8%     +80    __TEXT,__const
  +4.3%     +32  +4.3%     +32    Rebase Info
  +1.4%     +32  +1.4%     +32    __DATA,__objc_selrefs
  +2.6%     +24  +2.6%     +24    Function Start Addresses
  +0.8%     +24  +0.8%     +24    Lazy Binding Info
  +4.7%     +19  +4.7%     +19    __TEXT,__objc_classname
  -2.0%     -88  -2.0%     -88    [__DATA_CONST]
  -9.0% -2.90Ki  -9.0% -2.90Ki    [__TEXT]
  [ = ]       0 -77.2% -6.16Ki    [__LINKEDIT]
  +4.9% +22.2Ki  +3.4% +16.0Ki    TOTAL

Generated by 🚫 Danger