Akka.Actor: tuck all scheduled Tell messages into IScheduledMsg envelope

[Aaronontheweb]

Changes

Adds some overhead to actor message processing (minor), but makes it possible for us to detect scheduled messages and filter them out from telemetry. Doesn't need to ship as part of v1.5 - can be done after.

Checklist

For significant changes, please ensure that the following have been completed (delete if not relevant):

• This change follows the Akka.NET API Compatibility Guidelines.
• This change follows the Akka.NET Wire Compatibility Guidelines.
• I have reviewed my own pull request.
• Changes in public API reviewed, if any.

[Aaronontheweb]

Motivation for this is primarily telemetry - making it easier to determine what is and is not a scheduled message. I'd probably rather make this a 1.5.1 thing though since I think I need more time to let it cook...

[Aaronontheweb]

@Arkatufus if the tests look good, next thing we need to run are benchmarks - InMemoryPingPong and RemotePingPong, since this touches the hotpath in the ActorCell.

[Arkatufus]

RemotePingPong Benchmark

Values are a median of 5 runs

New Code

Num clients	Total [msg]	Msg/sec	Total [ms]	Start Threads	End Threads
1	200000	138409	1445.25	131	158
5	1000000	611996	1634.23	166	174
10	2000000	676590	2956.19	177	181
15	3000000	666815	4499.7	189	193
20	4000000	664784	6017.72	207	199
25	5000000	670152	7461.99	199	191
30	6000000	665042	9022.91	199	181

Dev Branch

Num clients	Total [msg]	Msg/sec	Total [ms]	Start Threads	End Threads
1	200000	131927	1516.93	132	159
5	1000000	608273	1644.59	166	171
10	2000000	674082	2967.81	179	180
15	3000000	669942	4478.95	189	200
20	4000000	664452	6020.36	207	219
25	5000000	670961	7452.86	207	191
30	6000000	668748	8972.71	200	182

[Arkatufus]

Akka.Benchmark PingPongBenchmarks Result

Dev Branch

Method	Mean	Error	StdDev	Req/sec
Actor_ping_pong_single_pair_in_memory	384.2 ns	5.89 ns	17.38 ns	2,603,016.96

New Code

Method	Mean	Error	StdDev	Req/sec
Actor_ping_pong_single_pair_in_memory	381.9 ns	3.95 ns	11.64 ns	2,618,609.51

[Aaronontheweb]

No negative performance impact - that's good.

[Aaronontheweb]

Left some suggestions

[Aaronontheweb]

LGTM

[Aaronontheweb]

Fixing up some bad test changes - looks fine