Fix breakpoint_manager not using its own retq finding function

Looks like there's a lot of duplicated code for the refactor
to clean up.

Finally, instead of having to hardcode a constant, the length of
the code to disassemble in search of a RET instruction is discovered
at runtime.

Increased the range of instructions a RET is searched for.
TODO: more robust error handling, see if there's a way to search up to a function's end via the GDB API

Also update PDFMainBreakpoint after moving set_hammer_retq_breakpoints() to BreakpointManager

Remove finished TODO entry

- final names TBD
- TODO: GDB commands for highest per-arena and total mem use