(Docs) Fix MLock constructor to not hang on first call to 'acquire'

[sloshy]

I think a coworker of mine found a bug in one of the MVar examples. https://typelevel.org/cats-effect/concurrency/mvar.html#use-case-asynchronous-lock-binary-semaphore-mutex

When you make a new MLock, it isn't initialized, so if you call acquire you will just be waiting forever. It basically starts in an already pre-acquired state which might be confusing. I replaced the MVar[IO].empty[Unit] call with a call to MVar[IO].of(()) which should work.

[codecov-io]

Codecov Report

Merging #597 into master will decrease coverage by 0.23%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master     #597      +/-   ##
==========================================
- Coverage   89.22%   88.99%   -0.24%     
==========================================
  Files          68       68              
  Lines        2144     2144              
  Branches      108      108              
==========================================
- Hits         1913     1908       -5     
- Misses        231      236       +5