be consistent about our defintion of main

[ppannuto]

Closes #338.

This mostly implements the discussion from the issue thread, it:

• marks our _start in crt1 as weak
• overrides _start in tests where access memory regions is nice
  • I view this as somewhat niche and a non-binding interface for now
• defines that standard main prototype as int main(void)

In the original issue thread, we'd suggested a prototype of void main(void),
as Tock currently ignores the return value of processes. Unfortuntely, -Wall
turns on -Wmain by default, which results in

main.c:6:6: warning: return type of 'main' is not 'int' [-Wmain]
 void main(void) {
      ^~~~

while we could explicitly turn this off (-Wno-main), my current feeling is
that writing int main is pretty stock for C, and having a return value from a
process may be useful someday (i.e. returning 0 means this process meant to
exit). Sure, most things probably won't ever return, but I don't think it hurts
to stick with C convention here and let things return a value.

[alevy]

Something to the effect of "returning from main with a value other than 0 is currently undefined"?

[alevy]

@ppannuto merge/squash at will