PowerPC, Fix-test-gdb.base-store.exp

[Carl Love <cel@us.ibm.com>]

GDB maintainers:

The gdb.base/store.exp test has five failures on PowerPC.  Part of the
issue is the PowerPC DWARF mapping is not correct on Linux.  The first
patch in this series fixes the PowerPC DWARF mapping.  Once the mapping
was fixed another issue was found with how the signal handlers were
handled on PowerPC.  The issue is on PowerPC the sequence of events for
handling signals is a little different on PowerPC from most other
architectures. 

The first patch fixes the PowerPC DWARF register mapping and the signal
handling issue on PowerPC.  These changes fix three of the five
failures on the store.exp test.

The second patch fixes the remaining issue for the store.exp test.  The
remaining issues are due to handling the 128-bit floating point values.
The 64-bit floating point values are stored in the floating point
registers while the 128-bit floating point values are stored in a VSR. 
The DWARF uses the floating point register numbers for both the 64-bit
and 128-bit floating point values.  The second patch adds a routine to
check if a value is a 128-bit floating point value, if so it then
adjusts the register mapping to point to the VSR registers instead of
the 64-bit floating point registers.

The series of two patches have been tested on PowerPC LE and BE.  The
patches have also been tested on Intel X86-64.  No regression failures
was found.

Please let me know if these patches are acceptable for mainline. 
Thanks.

                             Carl