Avoid null pointer dereference for partial configuration settings

[RyanHeule]

Fixes #13864.

I first looked at how other extruder configuration options were validated and/or normalized. I found them at:

PrusaSlicer/src/libslic3r/PrintObject.cpp

Lines 2598 to 2599 in 5c7888b

	clamp_exturder_to_default(config.support_material_extruder, num_extruders);
	clamp_exturder_to_default(config.support_material_interface_extruder, num_extruders);

and:

PrusaSlicer/src/libslic3r/PrintObject.cpp

Lines 2650 to 2652 in 5c7888b

	clamp_exturder_to_default(config.infill_extruder, num_extruders);
	clamp_exturder_to_default(config.perimeter_extruder, num_extruders);
	clamp_exturder_to_default(config.solid_infill_extruder, num_extruders);

Upon further analysis, I determined that wipe_tower_extruder validation does not really fit in either spot -- as far as I can tell, the wipe_tower_extruder can only be set at the print level, whereas the others can be set for individual objects (for supports) or even regions of an object (for the rest).

I considered adding another call of clamp_exturder_to_default [sic] in several other locations, such as within Print::validate(std::vector<std::string>*) or Print::apply(const Model &, DynamicPrintConfig). Quite possibly one of these spots could have worked, however due to my unfamiliarity with the code base, I was concerned that I may accidentally introduce other bugs. I also felt a large amount of testing would be required to validate such a change. For instance, I was not sure at what point it is no longer safe to change configuration (i.e. updating wipe_tower_extruder to a valid value) due to other logic already relying on it.

After a lot of deliberation, it seemed to me that DynamicPrintConfig::normalize_fdm() could work after all (despite my initial concerns expressed in #13864). I examined where it is called, and I found it is used in several stages of processing configuration -- for example, for each file provided by a --load command line option, for presets, for default FFF or SLA print options, and so on.

Even though early calls to DynamicPrintConfig::normalize_fdm() do not necessarily have full configuration (i.e. the original bug), for sure later calls do. And it seemed to me that someone had already tested wipe_tower_extruder validation at this location (just not without nozzle_diameter in the same file) to prove that this could work.

I did build this and do some very basic testing (a few command line calls, including what I was originally doing when I encountered the segmentation fault in the first place).