Update CMakeLists.txt

Remove GMP from knumber_float implementation and mark MPFR as required.

Ping

In D21495#474048, @cfeck wrote:

Evan, any objections? I would like to commit it to master branch.

Marcelo, please add a comment to bug 407318 with a link to this phabricator page. The mail address is used for the author attribution.

Fix binary precision

Remove duplicated #define HAVE_LONG_DOUBLE 1 in config-kcalc.h.cmake

In D21495#472200, @teran wrote:

In D21495#472071, @cfeck wrote:

Adding Evan since he knows best what prevented the use of MPFR when the code was initially added.

Thanks.

Honestly, it was disabled by default just because it was new and I didn't want any difference in behavior seemingly out of nowhere. The code is probably pretty good, though i expect some regression tests will have different results due to the difference in precision.

Mark MPFR as recommended

That is true.
I will consider taking a look at KPasswordLineEdit implementation.
Thanks a lot!