gdb/python: eval_python_command returns result of the evaluation

A previous commit [1] centralized the Python code evaluation in the
helper function eval_python_command(). However, a missed case in
varobj_set_visualizer() requires not only the exit status but the
result of the evaluation too.

This patch updates eval_python_command() to return the result of
the evaluation, or NULL if an error occurred. It also adjusts the
existing callers accordingly. Finally, it replaces the use of
PyRun_String() in varobj_set_visualizer() with a call to
eval_python_command().

[1]: 264a8a2236e8aa64b333a69e42a55ff8c0844f6e

Bug: https://sourceware.org/bugzilla/show_bug.cgi?id=23830

Approved-By: Tom Tromey <tom@tromey.com>