Check for NULL from proj_as_proj_string in proj_factors derived CRS handling

[ClayWarren]

Motivation

• A code path in proj_factors() for derived projected CRS constructed a std::string from the return value of proj_as_proj_string() without checking for NULL, which can occur when an object is not exportable or on export error and causes undefined behavior/crash.
• The change prevents denial-of-service / crash scenarios when processing crafted or non-exportable CRS definitions while preserving existing functionality.

Description

• Add a NULL check for the result of proj_as_proj_string() and return safely (after cleaning up) if export fails, avoiding construction of std::string from NULL in the derived projected CRS branch. (modified src/factors.cpp).
• When rebuilding an operation from the trimmed PROJ string, check the result of proj_create() and return safely on failure instead of relying only on assert() (modified src/factors.cpp).
• Add a regression unit test that constructs a derived projected CRS that cannot be exported to a PROJ string and asserts that proj_factors() returns an error condition rather than crashing (added test in test/unit/gie_self_tests.cpp).

Testing

• Configured with CMake (-DCMAKE_BUILD_TYPE=Release) successfully in this environment. (succeeded)

• Built the project with cmake --build successfully and produced binaries. (succeeded)

• Exercised the produced proj binary with a simple projection invocation (./build/bin/proj +proj=merc +ellps=WGS84) which produced expected numeric output. (succeeded)

• AI/LLM (Codex Security in research preview) supported my development of this PR

• Tests added

• Added clear title that can be used to generate release notes

[rouault]

@ClayWarren Why did you delete this PR ? This looked good to me

[ClayWarren]

I deleted my fork, I'm still learning.