[lmi] MinGW-w64 roundl() anomaly

[Greg Chicares]

'round_test.cpp' doesn't pretend to be a comprehensive validation suite
for the round() family, but it always succeeded with mingw.org's RTL
supplement (libmingwex), yet it fails with MinGW-w64. AFAICS, the only
problem it detects is
  roundl(0.499999999999999999973)
which I believe should always return 0.0L regardless of the current
hardware rounding mode. It seems to do that iff the current mode is
upward rounding--not if it's downward, toward zero, or to nearest.

Because this function is supposed to round to nearest always, I'd
expect its implementation to start by setting that mode, so an error
like this seems strange; but perhaps it's easy to fix.

Vadim, would you be interested in verifying my analysis and, if it's
correct, following up with the MinGW-w64 people to get this fixed?