Summary:
No functional changes except in panic messages.

These functions were identical except for
- x87 operations in genCCall32
- the fallback to genCCall32'/64'
- "32" vs "64" in panic messages (one case was wrong!)
- minor syntactic or otherwise non-functional differences.

Test Plan:
Ran "validate --no-dph --slow" before and after the change.
Only differences were two tests that failed before the change but not after,
further investigation revealed that those tests are in fact erratic.

Reviewers: simonmar, austin

Reviewed By: austin

Subscribers: phaskell, simonmar, relrod, ezyang, carter

Differential Revision: https://phabricator.haskell.org/D139