[stdlib] Stop swapping self in Array.withUnsafeMutableBufferPointer

[lorentey]

Array.withUnsafeMutableBufferPointer swaps out self's buffer for an empty array for the duration of its closure call, to prevent the buffer getting released by the closure.

This code predates the introduction of the law of exclusivity, and I think we don't need it anymore.

Removing the swap gets rid of the need to create an empty array, which eliminates a pair of swift_retain/release calls for the dummy value. This can make a measurable difference in performance sensitive code.

[lorentey]

@swift-ci test

[lorentey]

@swift-ci benchmark

[lorentey]

Cc @aschwaighofer, @eeckstein -- would this be an okay change? IIUC, the law of exclusivity prevents body from accessing or escaping self in any way. (Can we trust exclusivity checking to catch mistakes?)

[swift-ci]

Performance (x86_64): -O

Regression	OLD	NEW	DELTA	RATIO
ArrayAppendOptionals	1130	2080	+84.1%	0.54x (?)
DictionaryOfAnyHashableStrings_insert	2744	5040	+83.7%	0.54x
Breadcrumbs.MutatedUTF16ToIdx.ASCII	3	4	+33.3%	0.75x
Set.isDisjoint.Int25	241	309	+28.2%	0.78x (?)
Set.isDisjoint.Int50	241	303	+25.7%	0.80x (?)
DictionaryKeysContainsNative	19	23	+21.1%	0.83x (?)
ObjectiveCBridgeStubFromNSDateRef	3930	4540	+15.5%	0.87x (?)
String.data.Medium	90	101	+12.2%	0.89x (?)
Improvement	OLD	NEW	DELTA	RATIO
CSVParsing.Scalar	162	151	-6.8%	1.07x (?)

Code size: -O

Improvement	OLD	NEW	DELTA	RATIO
SortIntPyramids.o	9136	9040	-1.1%	1.01x

Performance (x86_64): -Osize

Regression	OLD	NEW	DELTA	RATIO
DictionaryKeysContainsNative	24	27	+12.5%	0.89x (?)
Improvement	OLD	NEW	DELTA	RATIO
ArrayAppendGenericStructs	2080	1130	-45.7%	1.84x (?)

Code size: -Osize

Performance (x86_64): -Onone

Regression	OLD	NEW	DELTA	RATIO
ClassArrayGetter2	4260	4650	+9.2%	0.92x (?)

Code size: -swiftlibs

How to read the data

The tables contain differences in performance which are larger than 8% and differences in code size which are larger than 1%.

If you see any unexpected regressions, you should consider fixing the
regressions before you merge the PR.

Noise: Sometimes the performance results (not code size!) contain false
alarms. Unexpected regressions which are marked with '(?)' are probably noise.
If you see regressions which you cannot explain you can try to run the
benchmarks again. If regressions still show up, please consult with the
performance team (@eeckstein).

Hardware Overview

  Model Name: Mac Pro
  Model Identifier: MacPro6,1
  Processor Name: 8-Core Intel Xeon E5
  Processor Speed: 3 GHz
  Number of Processors: 1
  Total Number of Cores: 8
  L2 Cache (per Core): 256 KB
  L3 Cache: 25 MB
  Memory: 64 GB

[eeckstein]

should be fine.
LGTM