Faster Processor resets

[milseman]

Make processor reset faster by tracking dirty registers
and only clearing Arrays that are non-empty.

Built on top of #773

[milseman]

rdar://137810309

[milseman]

Perf vs #773:

Comparing against saved benchmark result persist_state
=== Regressions ======================================================================
- EmailLookaheadList_Scalar               4.11ms	4.06ms	49.3µs		1.2%
- MACAddress_Scalar                       2.42ms	2.37ms	45.7µs		1.9%
- IPv6Address                             2.62ms	2.57ms	42.3µs		1.6%
=== Improvements =====================================================================
- CompilerMessages_All_Scalar             71ms	74.7ms	-3.68ms		-4.9%
- EmailRFCNoMatches_All                   48.9ms	51.9ms	-2.97ms		-5.7%
- EmailRFCNoMatches_All_Scalar            44.9ms	47.6ms	-2.69ms		-5.7%
- EmailLookaheadNoMatches_All             22.3ms	24.7ms	-2.38ms		-9.6%
- BasicRangeCCC_All_Scalar                4.04ms	6.36ms	-2.32ms		-36.5%
- BasicRangeCCC_All                       4.06ms	6.33ms	-2.27ms		-35.9%
- CaseInsensitiveCCC_All_Scalar           4.28ms	6.54ms	-2.26ms		-34.6%
- CaseInsensitiveCCC_All                  4.27ms	6.53ms	-2.26ms		-34.6%
- DiceRollsInText_All                     30.1ms	32.4ms	-2.25ms		-7.0%
- BasicCCC_All_Scalar                     3.91ms	6.11ms	-2.2ms		-36.0%
- DiceRollsInText_All_Scalar              28ms	30.1ms	-2.04ms		-6.8%
- BasicCCC_All                            3.93ms	5.89ms	-1.96ms		-33.3%
- NotFound_All                            4.82ms	6.61ms	-1.79ms		-27.1%
- LiteralSearchNotFound_All               3.93ms	5.56ms	-1.63ms		-29.3%
- EmailLookaheadNoMatches_All_Scalar      20.2ms	21.9ms	-1.63ms		-7.4%
- BasicBuiltinCharacterClass_All          4.03ms	5.66ms	-1.63ms		-28.7%
- SubtractionCCC_All_Scalar               6.31ms	7.93ms	-1.62ms		-20.4%
- IntersectionCCC_All                     5.95ms	7.56ms	-1.61ms		-21.3%
- IntersectionCCC_All_Scalar              5.99ms	7.58ms	-1.6ms		-21.1%
- SubtractionCCC_All                      6.31ms	7.9ms	-1.59ms		-20.1%
- LiteralSearch_All                       4.12ms	5.7ms	-1.58ms		-27.8%
- Numbers_All                             3.42ms	4.92ms	-1.5ms		-30.4%
- NotFound_All_Scalar                     3.71ms	5.2ms	-1.49ms		-28.7%
- LiteralSearch_All_Scalar                3.28ms	4.76ms	-1.48ms		-31.0%
- LiteralSearchNotFound_All_Scalar        3.15ms	4.6ms	-1.45ms		-31.5%
- symDiffCCC_All_Scalar                   18.1ms	19.5ms	-1.38ms		-7.1%
- symDiffCCC_All                          18.1ms	19.5ms	-1.36ms		-7.0%
- Numbers_All_Scalar                      2.63ms	3.96ms	-1.33ms		-33.6%
- HangulSyllable_All                      3.92ms	5.21ms	-1.29ms		-24.7%
- BasicBuiltinCharacterClass_All_Scalar   3.05ms	4.33ms	-1.28ms		-29.5%
- EmojiRegex_All                          125ms	126ms	-1.24ms		-1.0%
- HangulSyllable_All_Scalar               3.18ms	4.34ms	-1.15ms		-26.6%
- CompilerMessages_All                    90.1ms	91.2ms	-1.14ms		-1.2%
- EmailRFC_All                            32.8ms	33.8ms	-958µs		-2.8%
- EmojiRegex_All_Scalar                   43.7ms	44.7ms	-927µs		-2.1%
- HangulSyllable_First                    1.87ms	2.79ms	-919µs		-33.0%
- HangulSyllable_First_Scalar             1.49ms	2.24ms	-753µs		-33.6%
- EmailRFC_All_Scalar                     31.6ms	32.3ms	-698µs		-2.2%
- Words_All_Scalar                        5.2ms	5.84ms	-644µs		-11.0%
- Css_All_Scalar                          2.17ms	2.8ms	-636µs		-22.7%
- Words_All                               5.59ms	6.18ms	-591µs		-9.6%
- Css_All                                 2.55ms	3.11ms	-560µs		-18.0%
- InvertedCCC_All_Scalar                  6.79ms	7.21ms	-418µs		-5.8%
- InvertedCCC_All                         6.88ms	7.17ms	-295µs		-4.1%
- EmailLookahead_All_Scalar               17.7ms	18ms	-244µs		-1.4%
- URLWithWordBoundaries_All_SimpleWordBoundaries 674µs	900µs	-226µs		-25.1%
- ReluctantQuantWithTerminal_Whole        5.92ms	6.13ms	-215µs		-3.5%
- EmailBuiltinCharacterClass_All          9.21ms	9.41ms	-202µs		-2.1%
- URLWithWordBoundaries_All_Scalar        3.45ms	3.64ms	-195µs		-5.4%
- EmailLookahead_All                      20ms	20.2ms	-176µs		-0.9%
- EmailBuiltinCharacterClass_All_Scalar   8.97ms	9.13ms	-163µs		-1.8%
- AnchoredNotFound_First                  9.81ms	9.96ms	-152µs		-1.5%
- AnchoredNotFound_First_Scalar           5.91ms	6.05ms	-140µs		-2.3%
- ReluctantQuantWithTerminal_Whole_Scalar 5.92ms	6.04ms	-116µs		-1.9%
- ReluctantQuant_Whole                    9.84ms	9.93ms	-98.8µs		-1.0%
- Lines_All                               930µs	1.03ms	-95.5µs		-9.3%
- DiceNotation_Scalar                     4.55ms	4.64ms	-85.7µs		-1.8%
- Lines_All_Scalar                        888µs	957µs	-68.7µs		-7.2%
- GraphemeBreakNoCap_All                  1.78ms	1.84ms	-58.7µs		-3.2%
- AnchoredNotFound_All_Scalar             5.93ms	5.99ms	-54.9µs		-0.9%
- GraphemeBreakNoCap_All_Scalar           1.67ms	1.7ms	-29.1µs		-1.7%
- EagarQuantWithTerminal_Whole_Scalar     443µs	454µs	-10.8µs		-2.4%

[milseman]

@swift-ci please test

[natecook1000]

Need to reset isDirty here, as well.

[milseman]

Shoot, let's see if that affects numbers...

[milseman]

Oh, you mean we can set isDirty = false outside of the guard statement? Right, that will actually be an improvement.

[milseman]

Doing that change doesn't affect correctness, just allows us to reset less often. This nets us 10-15% in the Hangul benchmarks! I'll include this improvement in an up-coming PR.