Refactor Mirror to reduce metadata allocation

[airspeedswift]

Mirror uses AnyCollection to type erase the children between different custom mirror implementations and it's own default implementation. It was previously creating an AnyCollection to store the children on creation, and this was quite costly both in object and metadata creation.

This switches to an internal Either type to store the default children implementation, and only resorts to AnyCollection for custom implementations. This avoids the box allocation if you don't access the children. Unfortunately Mirror.children: AnyCollection<Any> is ABI, but internally the standard library can take a faster path.

[airspeedswift]

@swift-ci please benchmark

[airspeedswift]

@swift-ci please test

[swift-ci]

Performance: -O

Regression	OLD	NEW	DELTA	RATIO
Set.isSubset.Seq.Empty.Int	76	84	+10.5%	0.90x (?)
PrefixArrayLazy	13	14	+7.7%	0.93x (?)
Improvement	OLD	NEW	DELTA	RATIO
EqualSubstringSubstring	30	22	-26.7%	1.36x
LessSubstringSubstring	30	22	-26.7%	1.36x (?)
EqualSubstringSubstringGenericEquatable	30	22	-26.7%	1.36x
EqualSubstringString	30	22	-26.7%	1.36x
LessSubstringSubstringGenericComparable	30	22	-26.7%	1.36x
EqualStringSubstring	30	23	-23.3%	1.30x (?)
UTF8Decode_InitFromCustom_contiguous	165	144	-12.7%	1.15x (?)
StringUTF16Builder	240	210	-12.5%	1.14x
UTF8Decode_InitDecoding	163	143	-12.3%	1.14x (?)
StringComparison_longSharedPrefix	358	321	-10.3%	1.12x (?)
DistinctClassFieldAccesses	199	183	-8.0%	1.09x (?)
Set.isStrictSubset.Int.Empty	51	47	-7.8%	1.09x (?)
AngryPhonebook.Armenian	171	158	-7.6%	1.08x (?)

Code size: -O

Performance: -Osize

Regression	OLD	NEW	DELTA	RATIO
StringHasPrefixAscii	1170	1310	+12.0%	0.89x (?)
Improvement	OLD	NEW	DELTA	RATIO
EqualSubstringSubstring	30	22	-26.7%	1.36x
EqualSubstringSubstringGenericEquatable	30	22	-26.7%	1.36x
EqualSubstringString	30	22	-26.7%	1.36x
LessSubstringSubstring	30	23	-23.3%	1.30x
EqualStringSubstring	30	23	-23.3%	1.30x (?)
LessSubstringSubstringGenericComparable	30	23	-23.3%	1.30x
UTF8Decode_InitDecoding	164	143	-12.8%	1.15x (?)
UTF8Decode_InitFromCustom_contiguous	163	144	-11.7%	1.13x (?)
StringComparison_longSharedPrefix	359	324	-9.7%	1.11x (?)
StringUTF16Builder	230	210	-8.7%	1.10x (?)
AngryPhonebook.Cyrillic	183	169	-7.7%	1.08x (?)
AngryPhonebook.Armenian	171	158	-7.6%	1.08x (?)

Code size: -Osize

Performance: -Onone

Regression	OLD	NEW	DELTA	RATIO
StringWalk	2480	2920	+17.7%	0.85x (?)
Improvement	OLD	NEW	DELTA	RATIO
LessSubstringSubstring	35	27	-22.9%	1.30x
EqualSubstringSubstring	34	27	-20.6%	1.26x
EqualSubstringSubstringGenericEquatable	34	27	-20.6%	1.26x (?)
LessSubstringSubstringGenericComparable	34	27	-20.6%	1.26x (?)
EqualStringSubstring	35	28	-20.0%	1.25x
EqualSubstringString	35	28	-20.0%	1.25x
UTF8Decode_InitFromCustom_contiguous	175	153	-12.6%	1.14x (?)
UTF8Decode_InitDecoding	170	152	-10.6%	1.12x (?)
AngryPhonebook.Armenian	172	158	-8.1%	1.09x (?)
AngryPhonebook.Cyrillic	183	170	-7.1%	1.08x (?)

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 mini
  Model Identifier: Macmini8,1
  Processor Name: 6-Core Intel Core i7
  Processor Speed: 3.2 GHz
  Number of Processors: 1
  Total Number of Cores: 6
  L2 Cache (per Core): 256 KB
  L3 Cache: 12 MB
  Memory: 64 GB

[airspeedswift]

@swift-ci please test

[swift-ci]

Build failed
Swift Test Linux Platform
Git Sha - fc20d80

[swift-ci]

Build failed
Swift Test OS X Platform
Git Sha - fc20d80

[airspeedswift]

@swift-ci please test

[swift-ci]

Build failed
Swift Test Linux Platform
Git Sha - b1105ed

[airspeedswift]

@swift-ci smoke test linux

[AnthonyLatsis]

I think we can drop T by expressing this as Right == AnyCollection<C.Element>