Fix gcc -fsanitize detected problems

[BsAtHome]

An issue was detected, while doing unrelated stuff, that required a deeper dive. A sanitizer switch to gcc/g++ revealed more issues when compiled using -fsanitize=undefined,bool,float-cast-overflow (see also #421).

The changes include:

• Get rid of static char buffers in shcom and use std::string instead. This has effect on several dependent files, which were all adapted and already had FIXME markers at those places.
• Fix linuxcnc python module where the ini filename storage was changed to std::string previously. However, the python object structure has no C++ constructor semantics. The string must be separately allocated and cannot be a "normal" member. Otherwise it would not be properly initialized.
• Several off-by-one compiler messages in char buffers were fixed.
• Guard rcs_print() when a NULL format argument is passed.
• Compiling with the sanitizer tripped some stack-frame size warnings. These were slightly increased.
• A char buffer in halcmd was indexed with -1 on continuation search causing an underflow. That eventuality is now guarded against.

There were also unrelated boost::python problems with "maybe-uninitialized" warnings on value extract. However, it was determined that these were false positives (see also #3931).

It may be prudent to run the code base with other -fsanitize options and see the fallout.