Use Stopwatch.GetTimestamp on MacOS

[bitfaster]

Stopwatch.GetTimestamp is slightly faster than Environment.TickCount64 on MacOS, therefore use the faster implementation to get a small optimization.

This is the time benchmark result from Github actions running on MacOS:

This is the GetOrAdd bench (baseline v2.5.0):

BenchmarkDotNet v0.13.12, macOS Sonoma 14.4.1 (23E224) [Darwin 23.4.0]
Apple M1 (Virtual), 1 CPU, 3 logical and 3 physical cores
.NET SDK 8.0.204
  [Host]   : .NET 6.0.29 (6.0.2924.17105), Arm64 RyuJIT AdvSIMD
  .NET 6.0 : .NET 6.0.29 (6.0.2924.17105), Arm64 RyuJIT AdvSIMD

[coveralls]

coverage: 99.318% (+0.06%) from 99.263%
when pulling 6370b8d on users/alexpeck/macostime
into aadf0c7 on main.

[bitfaster]

On linux gettimeofday is likely fastest (and is a portable POSIX1 call):
https://stackoverflow.com/questions/48609413/fastest-way-to-get-a-timestamp

Can this be p/invoked on linux?
https://stackoverflow.com/questions/38202027/p-invoke-in-net-core-with-linux

Second answer here suggests it will also work on OSX:
https://stackoverflow.com/questions/5167269/clock-gettime-alternative-in-mac-os-x

P/Invoke docs:
https://learn.microsoft.com/en-us/dotnet/standard/native-interop/pinvoke

[bitfaster]

There is a repeatable very slight gain, but it is close to noise.

BenchmarkDotNet v0.13.12, macOS Sonoma 14.1.1 (23B81) [Darwin 23.1.0]
Apple M2, 1 CPU, 8 logical and 8 physical cores
.NET SDK 8.0.100
  [Host]   : .NET 6.0.29 (6.0.2924.17105), Arm64 RyuJIT AdvSIMD
  .NET 6.0 : .NET 6.0.29 (6.0.2924.17105), Arm64 RyuJIT AdvSIMD
Job=.NET 6.0  Runtime=.NET 6.0  

Method	Mean	Error	StdDev	Ratio
DateTimeUtcNow	15.135 ns	0.0100 ns	0.0078 ns	1.00
DateTimeOffsetUtcNow	20.768 ns	0.0170 ns	0.0142 ns	1.37
EnvironmentTickCount	10.624 ns	0.0036 ns	0.0028 ns	0.70
EnvironmentTickCount64	13.357 ns	0.0389 ns	0.0345 ns	0.88
PInvokeTickCount64	NA	NA	NA	?
StopWatchGetElapsed	12.174 ns	0.0024 ns	0.0019 ns	0.80
StopWatchGetTimestamp	11.585 ns	0.0018 ns	0.0014 ns	0.77
DurationSinceEpoch	9.115 ns	0.0089 ns	0.0074 ns	0.60

LRU bench (with fix)

Method	Mean	Error	StdDev	Ratio	Allocated
ConcurrentDictionary	3.914 ns	0.0508 ns	0.0475 ns	1.00	-
FastConcurrentLru	5.063 ns	0.0773 ns	0.0723 ns	1.29	-
ConcurrentLru	7.566 ns	0.1483 ns	0.1387 ns	1.93	-
AtomicFastLru	11.718 ns	0.0071 ns	0.0055 ns	3.01	-
FastConcurrentTLru	16.019 ns	0.0046 ns	0.0036 ns	4.11	-
FastConcLruAfterAccess	18.739 ns	0.0030 ns	0.0024 ns	4.81	-
FastConcLruAfter	20.636 ns	0.0063 ns	0.0049 ns	5.30	-
ConcurrentTLru	20.652 ns	0.0108 ns	0.0091 ns	5.30	-
ConcurrentLfu	16.370 ns	0.3445 ns	0.6300 ns	4.35	-
ClassicLru	20.232 ns	0.2783 ns	0.2603 ns	5.17	-
RuntimeMemoryCacheGet	58.372 ns	0.0678 ns	0.0530 ns	14.99	32 B
ExtensionsMemoryCacheGet	25.906 ns	0.0379 ns	0.0296 ns	6.65	24 B

v2.5.0 baseline

Method	Mean	Error	StdDev	Ratio	Allocated
ConcurrentDictionary	3.935 ns	0.0771 ns	0.0721 ns	1.00	-
FastConcurrentLru	5.013 ns	0.0246 ns	0.0206 ns	1.27	-
ConcurrentLru	7.608 ns	0.0991 ns	0.0927 ns	1.93	-
AtomicFastLru	11.790 ns	0.1040 ns	0.0973 ns	3.00	-
FastConcurrentTLru	16.610 ns	0.0064 ns	0.0054 ns	4.22	-
FastConcLruAfterAccess	18.822 ns	0.0069 ns	0.0058 ns	4.79	-
FastConcLruAfter	21.149 ns	0.0166 ns	0.0139 ns	5.38	-
ConcurrentTLru	21.449 ns	0.0067 ns	0.0052 ns	5.46	-
ConcurrentLfu	15.857 ns	0.0969 ns	0.0906 ns	4.03	-
ClassicLru	21.114 ns	0.0273 ns	0.0213 ns	5.37	-
RuntimeMemoryCacheGet	50.518 ns	0.3877 ns	0.3626 ns	12.84	32 B
ExtensionsMemoryCacheGet	25.940 ns	0.3420 ns	0.3199 ns	6.59	24 B