[CI] [C++] Turn optimizations off on AppVeyor

[asfimport]

Some AppVeyor entries build in release mode (for example MinGW, R). We should try to turn compiler optimizations off (e.g. by passing a CXXFLAGS) to make those entries faster.

Reporter: Antoine Pitrou / @pitrou

PRs and other links:

• GitHub Pull Request #6330

_{Note: This issue was originally created as ARROW-7669. Please see the migration documentation for further details.}

[asfimport]

Antoine Pitrou / @pitrou:
cc @kou @nealrichardson

[asfimport]

Neal Richardson / @nealrichardson:
For R I'm planning to take up ARROW-7675 after 0.16. In general we should probably migrate some/all jobs off of Appveyor since it's now our bottleneck.

[asfimport]

Antoine Pitrou / @pitrou:
While the build script would be different, that's a bit orthogonal still. Builds will take time on GHA as well (actually, perhaps even more, since GHA doesn't have sufficient caching for us).

[asfimport]

Neal Richardson / @nealrichardson:
You're right, it is orthogonal, though my thought was that using GHA and thus having more concurrent builds will be a bigger win.

[asfimport]

Antoine Pitrou / @pitrou:
Should not be a problem anymore now that we've moved R and MinGW builds to Github Actions.