proposal: "runtime/sync" package

[niaow]

In order to use interrupts to implement wakeup from async operations, we need some high-level synchronization primitives to be available to the user. In the #606 experiment, a condition variable was added to a new package called machine/sync, as can be seen here.

Some remaining questions for this are:

1. What primitives are needed for common use cases?
2. What should this package be called (machine/sync was the first thing that came to mind, but it isn't necessarily that great)?

[deadprogram]

What about just using aka actually implementing the standard sync? Do we need a separate package?

[niaow]

This is different, so yes we need a separate package.

[aykevl]

I would propose runtime/sync, or maybe even runtime/interrupt. The reason for putting it under the interrupt package is because I'd like to keep special cases for the compiler limited to the runtime package as much as possible.

[niaow]

In #1011, I went with runtime/sync as suggested

[deadprogram]

Now that #1142 has been merged, I think we can mark this issue as "next release". Is this correct @jaddr2line ?

[niaow]

No, I think this issue is more closely related to interrupt-safe condition variables which it seems are going to need to be held back to v0.15

[niaow]

#1172 dealt with this but I forgot about this issue.