-Wfloat-equal does not check for exact representation correctly

The following program demonstrates the issue:
```cpp
bool test(double d){
return d==1.100000000000000088817841970012523233890533447265625;
}
```
This generates -Wfloat-equal, but writing the equivalent value as a hexadecimal floating literal `0X1199999999999AP-52` instead of `1.100000000000000088817841970012523233890533447265625` causes no warning to be generated.

Start by compiling the supplied C++ reproducer with -Wfloat-equal and compare the decimal and hexadecimal spellings shown in the issue. Trace the warning decision for these two equivalent literals, then verify that the warning behavior is consistent for both forms.

Written by the indexing model from the issue text.