compiler-rt: fix f80 ceil/floor optimization

Our implementation did the classic add-sub rounding trick `(y = x +/- C =+ C - x)`
with `C = 1 / eps(T) = 2^(mantissa - 1)`. This approach only works for values whose
magnitude is below the rounding capacity of the constant. For a 64-bit mantissa
(like f80 has), `C = 2^63` only rounds for `|x| < 2^63`. Before we allowed this to
be ran on `e < bias + 64` aka `|x| < 2^64`. And because it isn't large enough,
we lose a bit to rounding.

For reference, the musl implementation does the same thing, using `mantissa - 1`:
https://git.musl-libc.org/cgit/musl/tree/src/math/ceill.c#n18
where `LDBL_MANT_DIG` is 64 for `long double` on x86.

This commit also combines the floor and ceil implementations into one generic one.

CMakeLists.txt

@@ -221,7 +221,6 @@ set(ZIG_STAGE2_SOURCES
     lib/compiler_rt/aulldiv.zig
     lib/compiler_rt/aullrem.zig
     lib/compiler_rt/bswap.zig
-    lib/compiler_rt/ceil.zig
     lib/compiler_rt/clear_cache.zig
     lib/compiler_rt/cmp.zig
     lib/compiler_rt/cmpdf2.zig
@@ -312,7 +311,7 @@ set(ZIG_STAGE2_SOURCES
     lib/compiler_rt/floatuntisf.zig
     lib/compiler_rt/floatuntitf.zig
     lib/compiler_rt/floatuntixf.zig
-    lib/compiler_rt/floor.zig
+    lib/compiler_rt/floor_ceil.zig
     lib/compiler_rt/fma.zig
     lib/compiler_rt/fmax.zig
     lib/compiler_rt/fmin.zig