Support Rocky Linux 9

[ninotarantino]

Get CML building and running on Rocky 9 in preparation for the Flight Sciences Lab (the runtime environment of most of our users) upgrading to Rocky 9 in September.

Issues resolved for Rocky 9:

• Somehow, a few cases in input files where a string was being assigned directly into a JEOD NamedItem were acceptable on Rocky 8 but not on Rocky 9. name = "string" → name.set_name("string")
• The StateInitialize::generate_random function explicitly casts the input seed to result_type to avoid potential signedness issues.
• Rewrote one of the stub unit tests to correctly catch messages and exceptions without leaking memory.
• Ignored a few more leak sources originating from code we can't control in LSAN outputs.
• Pre-addressed implicit enum to float conversions, which are deprecated in C++20. float_type = enum_type → float_type = static_cast<double>(enum_type)

Also addresses an issue in newer SWIG versions which arises when you inherit from a type but also include that type as a member of your class.

class Base {
public:
    Base(const Base& other) = delete;
    Base& operator=(const Base& other) = delete;
};

class Derived : public Base {
public:
    Derived(const Derived& other) = delete;
    Derived& operator=(const Derived& other) = delete;

    Base instance; /* (--) SWIG issues here */
};

Even though we've deleted the copy constructor and copy assignment operator for the base and derived classes, SWIG still generates assignment interface code for instance, which causes a compiler error. We're seeing this on Fedora 44 and macOS, but we may as well get ahead of fixing it now. The solution is to surround the instances of Base with

#ifdef SWIG
%immutable;
#endif
    Base instance;
#ifdef SWIG
%mutable;
#endif

The %immutable directive is a bit of a misnomer. It just prevents SWIG from generating set_X interface functions, it doesn't prevent calling member functions which mutate the type or setting data within that instance. So we're essentially just reinforcing what SWIG should already be recognizing, which is that the type is non-assignable.

Closes #31

[coveralls]

coverage: 90.223% (+0.009%) from 90.214% — support-rocky-9 into main

[ddj116]

Only did a quick review via my phone as I'm already on vacation! Looks good and see ya on 8/24!