Remove hack about the size of long doubles from DataExtractor

[tberghammer]

Summary

Remove hack about the size of long doubles from DataExtractor

The size of a long double was hardcoded in DataExtractor for x86 and
x86_64 architectures. This CL removes the hard coded values and use the
actual size based on the floating point semantics specified.

This CL depends on D8413

[clayborg]

Fine as long as clang no longer asserts and float values display correctly on x86_64 systems.

lldb/trunk/source/Core/DataExtractor.cpp

for (uint32_t count = 0; ValidOffset(offset) && count < item_count; ++count)
                                 if (GetAPInt (*this, &offset, item_byte_size, apint))
                                 {
                                     llvm::APFloat apfloat (ast->getFloatTypeSemantics(ast->DoubleTy), apint);
                                     apfloat.toString(sv, format_precision, format_max_padding);
                                 }
                             }
                             else if (item_bit_size == ast->getTypeSize(ast->LongDoubleTy))
                             {
-                                auto byte_size = item_byte_size;
                                 const auto &semantics = ast->getFloatTypeSemantics(ast->LongDoubleTy);
-                                if (&semantics == &llvm::APFloat::x87DoubleExtended)
-                                    byte_size = 10;
+                                const auto byte_size = (llvm::APFloat::getSizeInBits(semantics) + 7) / 8;
 
                                 llvm::APInt apint;
                                 if (GetAPInt(*this, &offset, byte_size, apint))
                                 {
                                     llvm::APFloat apfloat(semantics, apint);
                                     apfloat.toString(sv, format_precision, format_max_padding);
                                 }
                             }