Add command to scan mapping xrefs

Adds a repeat repeat_count attribute to GenericCommand that allows a
command to determine if it's being called again by the user just
pressing enter. This allows the command, for example to advance forward
through memory, in the case of hexdump.

Hexdump and Dereference now use this.

…sed section addresses (#339)

Add new gdb.Functions that can be used in gdb expressions. Added a set
of 'offset' functions that can be used to add an offset to its argument,
making it easy to get an address relative to a section.

…here the

warning is irrelevant is when debugging in qemu compat mode. We can safely
ignore that case if the qemu flag is set.

The format function doesn't ignore control codes, so we were padding
assuming they were visible.

In context_regs and 'registers' command, instead of colouring the value
of the register based on whether it changed since the last stop, colour
the name. This allows us to colour the value based on the section, as we
do with `deref`.

The ops for a `call` instruction may be something like the following.

call <func_name at filename.c:42>

The current regex will attempt to lookup the global symbol "func_name at
filename.c:42". Instead we should lookup just "func_name".

bufferize takes a function, but we were accidentally passing in the result of invoke, not invoke itself.

We missed an bug we introduced, because it only came up when a return
value that should have been a string but was not was was colorized.
Instead let's let the colour through, and strip it out after.

…cii value in `grep` command

Fixes

Minimize arch-specific code in commands

On ARM, AARCH64, and x86 we were not correctly comparing if two flags
were equal: Because we were comparing the masked value instead of
whether the bit was set, these would NEVER be true.

For example for `ge` on x86, the branch is taken is SF==OF, but we were
comparing  flags&(1<<7) == flags&(1<<11). Even if both flags were set,
we'd be comparing 0b10000000 with 0b100000000000, which would never be
true.