RNA: generate valid infinite float property default values

In `makesrna`, generate valid default values for float properties, when those
defaults are positive/negative infinity.

Blender would currently generate the strings `inff` and `-inff`, which are not
valid C++. This PR changes that to `std::numeric_limits<float>::infinity()` and
`-std::numeric_limits<float>::infinity()`.

Context: the layered animation system (see #113594) will have animation layers
with strips on them. It will have the concept of "infinite strips", where from
a user's perspective there aren't even any strips at all, and the layer seems
to directly contain the keys themselves. This is purely a UI distinction
though, where we simply won't draw the outline of infinite strips. Being able
to work with ±∞ as floating point values has clear advantages, as then the
`strip.frame_start` and `strip.frame_end` properties will always "tell the
truth".

This particular change is necessary to allow these values as defaults. It is
not intended that users will have property sliders for these properties when
their value is "infinite" (but rather a button 'make finite' or something along
those lines), so interaction with the UI code will likely be minimal.

Pull Request: https://projects.blender.org/blender/blender/pulls/113854

source/blender/makesrna/intern/makesrna.cc

@@ -646,6 +646,12 @@ static void rna_float_print(FILE *f, float num)
   else if ((fabsf(num) < float(INT64_MAX)) && (int64_t(num) == num)) {
     fprintf(f, "%.1ff", num);
   }
+  else if (num == std::numeric_limits<float>::infinity()) {
+    fprintf(f, "std::numeric_limits<float>::infinity()");
+  }
+  else if (num == -std::numeric_limits<float>::infinity()) {
+    fprintf(f, "-std::numeric_limits<float>::infinity()");
+  }
   else {
     fprintf(f, "%.10ff", num);
   }