Implement thread-local caching for Utf8JsonWriter in JsonSerializer.

[eiriktsarpalis]

Following recommendation in #69889. Records nice improvements in the allocation profile when serializing small-ish values.

cc @davidfowl

WriteJson<int>

Method	Job	Toolchain	Mean	Error	StdDev	Median	Min	Max	Ratio	MannWhitney(3%)	RatioSD	Gen 0	Allocated	Alloc Ratio
SerializeToString	Job-PLUXQS	main	120.17 ns	2.334 ns	2.184 ns	119.68 ns	116.27 ns	125.73 ns	1.00	Base	0.00	0.0182	184 B	1.00
SerializeToString	Job-YSJKRU	PR	123.69 ns	2.042 ns	1.705 ns	123.15 ns	120.87 ns	126.57 ns	1.03	Same	0.02	0.0030	32 B	0.17
SerializeToUtf8Bytes	Job-PLUXQS	main	105.70 ns	2.059 ns	1.826 ns	106.01 ns	102.20 ns	109.37 ns	1.00	Base	0.00	0.0181	184 B	1.00
SerializeToUtf8Bytes	Job-YSJKRU	PR	99.58 ns	2.058 ns	2.202 ns	98.58 ns	97.10 ns	103.77 ns	0.94	Faster	0.03	0.0028	32 B	0.17

WriteJson<LoginViewModel>

Method	Job	Branch	Mean	Error	StdDev	Median	Min	Max	Ratio	MannWhitney(3%)	RatioSD	Gen 0	Allocated	Alloc Ratio
SerializeToString	Job-PLUXQS	main	315.8 ns	10.84 ns	12.49 ns	314.0 ns	294.3 ns	338.3 ns	1.00	Base	0.00	0.0338	344 B	1.00
SerializeToString	Job-YSJKRU	PR	302.3 ns	4.02 ns	3.14 ns	302.8 ns	293.6 ns	306.6 ns	0.98	Same	0.03	0.0180	192 B	0.56
SerializeToUtf8Bytes	Job-PLUXQS	main	278.0 ns	5.09 ns	4.52 ns	279.4 ns	270.3 ns	284.2 ns	1.00	Base	0.00	0.0252	264 B	1.00
SerializeToUtf8Bytes	Job-YSJKRU	PR	282.4 ns	5.73 ns	6.37 ns	281.5 ns	273.0 ns	296.5 ns	1.02	Same	0.03	0.0110	112 B	0.42

WriteJson<Dictionary<string,string>> (Count == 100)

Method	Job	Toolchain	Mean	Error	StdDev	Median	Min	Max	Ratio	MannWhitney(3%)	RatioSD	Gen 0	Gen 1	Allocated	Alloc Ratio
SerializeToString	Job-PLUXQS	main	9.233 μs	0.1798 μs	0.1594 μs	9.188 μs	8.996 μs	9.466 μs	1.00	Base	0.00	1.1623	0.0375	12008 B	1.00
SerializeToString	Job-YSJKRU	PR	9.258 μs	0.2519 μs	0.2900 μs	9.243 μs	8.797 μs	9.957 μs	1.00	Same	0.04	1.1621	0.0363	11856 B	0.99
SerializeToUtf8Bytes	Job-PLUXQS	main	8.349 μs	0.0805 μs	0.0673 μs	8.352 μs	8.235 μs	8.448 μs	1.00	Base	0.00	0.5759	-	6096 B	1.00
SerializeToUtf8Bytes	Job-YSJKRU	PR	8.817 μs	0.4493 μs	0.5174 μs	8.516 μs	8.335 μs	10.016 μs	1.08	Same	0.07	0.5688	-	5944 B	0.98

Fix #69889.

[ghost]

Tagging subscribers to this area: @dotnet/area-system-text-json, @gregsdennis
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Following recommendation in #69889. Records nice improvements in the allocation profile when serializing small-ish values.

cc @davidfowl

WriteJson<int>

Method	Job	Toolchain	Mean	Error	StdDev	Median	Min	Max	Ratio	MannWhitney(3%)	RatioSD	Gen 0	Allocated	Alloc Ratio
SerializeToString	Job-PLUXQS	main	120.17 ns	2.334 ns	2.184 ns	119.68 ns	116.27 ns	125.73 ns	1.00	Base	0.00	0.0182	184 B	1.00
SerializeToString	Job-YSJKRU	PR	123.69 ns	2.042 ns	1.705 ns	123.15 ns	120.87 ns	126.57 ns	1.03	Same	0.02	0.0030	32 B	0.17
SerializeToUtf8Bytes	Job-PLUXQS	main	105.70 ns	2.059 ns	1.826 ns	106.01 ns	102.20 ns	109.37 ns	1.00	Base	0.00	0.0181	184 B	1.00
SerializeToUtf8Bytes	Job-YSJKRU	PR	99.58 ns	2.058 ns	2.202 ns	98.58 ns	97.10 ns	103.77 ns	0.94	Faster	0.03	0.0028	32 B	0.17

WriteJson<LoginViewModel>

Method	Job	Branch	Mean	Error	StdDev	Median	Min	Max	Ratio	MannWhitney(3%)	RatioSD	Gen 0	Allocated	Alloc Ratio
SerializeToString	Job-PLUXQS	main	315.8 ns	10.84 ns	12.49 ns	314.0 ns	294.3 ns	338.3 ns	1.00	Base	0.00	0.0338	344 B	1.00
SerializeToString	Job-YSJKRU	PR	302.3 ns	4.02 ns	3.14 ns	302.8 ns	293.6 ns	306.6 ns	0.98	Same	0.03	0.0180	192 B	0.56
SerializeToUtf8Bytes	Job-PLUXQS	main	278.0 ns	5.09 ns	4.52 ns	279.4 ns	270.3 ns	284.2 ns	1.00	Base	0.00	0.0252	264 B	1.00
SerializeToUtf8Bytes	Job-YSJKRU	PR	282.4 ns	5.73 ns	6.37 ns	281.5 ns	273.0 ns	296.5 ns	1.02	Same	0.03	0.0110	112 B	0.42

WriteJson<Dictionary<string,string>> (Count == 100)

Method	Job	Toolchain	Mean	Error	StdDev	Median	Min	Max	Ratio	MannWhitney(3%)	RatioSD	Gen 0	Gen 1	Allocated	Alloc Ratio
SerializeToString	Job-PLUXQS	main	9.233 μs	0.1798 μs	0.1594 μs	9.188 μs	8.996 μs	9.466 μs	1.00	Base	0.00	1.1623	0.0375	12008 B	1.00
SerializeToString	Job-YSJKRU	PR	9.258 μs	0.2519 μs	0.2900 μs	9.243 μs	8.797 μs	9.957 μs	1.00	Same	0.04	1.1621	0.0363	11856 B	0.99
SerializeToUtf8Bytes	Job-PLUXQS	main	8.349 μs	0.0805 μs	0.0673 μs	8.352 μs	8.235 μs	8.448 μs	1.00	Base	0.00	0.5759	-	6096 B	1.00
SerializeToUtf8Bytes	Job-YSJKRU	PR	8.817 μs	0.4493 μs	0.5174 μs	8.516 μs	8.335 μs	10.016 μs	1.08	Same	0.07	0.5688	-	5944 B	0.98

Fix #69889.

Author:	eiriktsarpalis
Assignees:	eiriktsarpalis
Labels:	area-System.Text.Json, tenet-performance
Milestone:	7.0.0