Refactor ErrorHandler

[MadVikingGod]

This change was made to enable ALL updates to the ErrorHandler to propagate to every captured GetErrorHandler(). This also was an attempt to make the error handler (and eventually the global logger) simpler. It had the side effect of also making these calls more performant.

Note: When first benchmarking these changes the resetting of the timers made both old and new benchmarks take longer, and occasionally crash. All the benchmark numbers that follow are with the StopTimer() and StartTimer() removed. I can post some before and after stats, but the net effect is additional allocation, because that's what the stop and start guarded, and faster run times, lower ns/op. I don't think the timing is accurate if you pause often within the test.

Note2: I removed one benchmark, because in the new method EVERYTHING is a delegate, so there is no difference between the 1 SetErrorHandler and multiple.

What I'm seeing performance-wise from this change is:

• A small performance degradation (16 ns) in calling the default logger.
• A similar performance increase when calling anything but the default.
• 1 fewer allocations for each SetErrorHandler is called.
• A large performance increase for GetErrorHandler, this is because it always returns the same location.
• A net performance increase in the general test.

$ benchstat old new
name                           old time/op    new time/op    delta
ErrorHandler-4                   1.98µs ± 5%    1.86µs ± 4%   -6.29%  (p=0.000 n=9+10)
GetDefaultErrorHandler-4         2.06ns ± 3%    0.78ns ± 8%  -62.03%  (p=0.000 n=9+10)
GetDelegatedErrorHandler-4       2.04ns ± 3%    0.92ns ± 6%  -55.01%  (p=0.000 n=10+9)
DefaultErrorHandlerHandle-4       289ns ± 2%     305ns ± 3%   +5.63%  (p=0.000 n=10+9)
DelegatedErrorHandlerHandle-4     313ns ± 2%     306ns ± 3%   -2.27%  (p=0.003 n=10+9)
SetErrorHandlerDelegation-4       293ns ± 4%     157ns ± 3%  -46.39%  (p=0.000 n=10+9)

name                           old alloc/op   new alloc/op   delta
ErrorHandler-4                     192B ± 0%      144B ± 0%  -25.00%  (p=0.000 n=10+10)
GetDefaultErrorHandler-4          0.00B          0.00B          ~     (all equal)
GetDelegatedErrorHandler-4        0.00B          0.00B          ~     (all equal)
DefaultErrorHandlerHandle-4       48.0B ± 0%     48.0B ± 0%     ~     (all equal)
DelegatedErrorHandlerHandle-4     48.0B ± 0%     48.0B ± 0%     ~     (all equal)
SetErrorHandlerDelegation-4        152B ± 0%      136B ± 0%  -10.53%  (p=0.000 n=10+10)

name                           old allocs/op  new allocs/op  delta
ErrorHandler-4                     9.00 ± 0%      6.00 ± 0%  -33.33%  (p=0.000 n=10+10)
GetDefaultErrorHandler-4           0.00           0.00          ~     (all equal)
GetDelegatedErrorHandler-4         0.00           0.00          ~     (all equal)
DefaultErrorHandlerHandle-4        1.00 ± 0%      1.00 ± 0%     ~     (all equal)
DelegatedErrorHandlerHandle-4      1.00 ± 0%      1.00 ± 0%     ~     (all equal)
SetErrorHandlerDelegation-4        5.00 ± 0%      4.00 ± 0%  -20.00%  (p=0.000 n=10+10)

[codecov]

Codecov Report

Merging #2424 (c351302) into main (a4f183b) will increase coverage by 0.0%.
The diff coverage is 100.0%.

@@          Coverage Diff          @@
##            main   #2424   +/-   ##
=====================================
  Coverage   75.3%   75.3%           
=====================================
  Files        173     173           
  Lines      11954   11948    -6     
=====================================
- Hits        9007    9003    -4     
+ Misses      2706    2704    -2     
  Partials     241     241           

Impacted Files	Coverage Δ
handler.go	100.0% <100.0%> (ø)
sdk/trace/batch_span_processor.go	84.4% <0.0%> (+1.0%)	⬆️

[MadVikingGod]

Skipping changelog as this is a non-user facing change.