Improve path search in Kestrel parsing

[BrennanConroy]

Replaces a manual loop with IndexOfAny which is much faster when the paths are longer than ~5 bytes.

The HttpParserBenchmark shows large improvements in the larger path cases and some minor regressions in the small paths. Ran twice since there is some noise in the runs.
Before:

Method	Mean	Error	StdDev	Op/s	Ratio	Gen 0	Gen 1	Gen 2	Allocated
PlaintextTechEmpower	79.16 ns	0.274 ns	0.243 ns	12,633,030.5	1.00	-	-	-	-
JsonTechEmpower	77.78 ns	0.283 ns	0.251 ns	12,857,132.1	0.98	-	-	-	-
LiveAspNet	140.69 ns	0.800 ns	0.710 ns	7,107,794.5	1.78	-	-	-	-
Unicode	248.81 ns	0.630 ns	0.589 ns	4,019,189.1	3.14	-	-	-	-
MultispanUnicodeHeader	348.31 ns	1.285 ns	1.202 ns	2,871,001.9	4.40	-	-	-	-

Method	Mean	Error	StdDev	Op/s	Ratio	Gen 0	Gen 1	Gen 2	Allocated
PlaintextTechEmpower	78.22 ns	0.261 ns	0.231 ns	12,785,243.6	1.00	-	-	-	-
JsonTechEmpower	75.88 ns	0.288 ns	0.270 ns	13,178,565.6	0.97	-	-	-	-
LiveAspNet	141.37 ns	0.845 ns	0.705 ns	7,073,883.6	1.81	-	-	-	-
Unicode	244.38 ns	0.743 ns	0.695 ns	4,092,016.3	3.12	-	-	-	-
MultispanUnicodeHeader	345.19 ns	1.052 ns	0.933 ns	2,896,948.9	4.41	-	-	-	-

---

After:

Method	Mean	Error	StdDev	Op/s	Ratio	RatioSD	Gen 0	Gen 1	Gen 2	Allocated
PlaintextTechEmpower	78.01 ns	0.379 ns	0.355 ns	12,818,186.3	1.00	0.00	-	-	-	-
JsonTechEmpower	77.80 ns	0.744 ns	0.696 ns	12,853,080.1	1.00	0.01	-	-	-	-
LiveAspNet	143.62 ns	0.187 ns	0.156 ns	6,962,883.2	1.84	0.01	-	-	-	-
Unicode	192.67 ns	0.426 ns	0.332 ns	5,190,248.7	2.47	0.01	-	-	-	-
MultispanUnicodeHeader	305.62 ns	2.585 ns	2.291 ns	3,272,037.9	3.92	0.04	-	-	-	-

Method	Mean	Error	StdDev	Op/s	Ratio	RatioSD	Gen 0	Gen 1	Gen 2	Allocated
PlaintextTechEmpower	76.55 ns	0.554 ns	0.518 ns	13,062,710.7	1.00	0.00	-	-	-	-
JsonTechEmpower	83.28 ns	0.534 ns	0.500 ns	12,007,535.9	1.09	0.01	-	-	-	-
LiveAspNet	142.14 ns	1.106 ns	1.034 ns	7,035,509.7	1.86	0.02	-	-	-	-
Unicode	196.96 ns	0.784 ns	0.655 ns	5,077,196.0	2.57	0.02	-	-	-	-
MultispanUnicodeHeader	307.96 ns	1.276 ns	1.131 ns	3,247,125.9	4.02	0.03	-	-	-	-

Also ran a more micro benchmark to show the difference between just the changed code:

Method	Path Length	Mean	Error	StdDev
Main	1	1.802 ns	0.0131 ns	0.0122 ns
PR	1	4.484 ns	0.0387 ns	0.0324 ns
Main	5	4.890 ns	0.0185 ns	0.0164 ns
PR	5	5.818 ns	0.0234 ns	0.0219 ns
Main	10	9.653 ns	0.0382 ns	0.0319 ns
PR	10	4.477 ns	0.0213 ns	0.0199 ns
Main	26	24.156 ns	0.0398 ns	0.0311 ns
PR	26	4.478 ns	0.0120 ns	0.0100 ns
Main	62	58.653 ns	0.4248 ns	0.3974 ns
PR	62	8.060 ns	0.0585 ns	0.0489 ns

[ghost]

Looks like this PR hasn't been active for some time and the codebase could have been changed in the meantime.
To make sure no conflicting changes have occurred, please rerun validation before merging. You can do this by leaving an /azp run comment here (requires commit rights), or by simply closing and reopening.

[BrennanConroy]

/azp run

[azure-pipelines]

Azure Pipelines successfully started running 3 pipeline(s).