Optimize depletion chain loading and material activity calculations

[paulromano]

Description

I recently found while profiling a script that analyzed activated material compositions from an R2S calculation that substantial time was being spent repeatedly resolving depletion-chain paths and parsing decay source distributions. This PR fixes this by loading the depletion chain once per Material.get_activity() call, utilizing Path.absolute() instead of Path.resolve(), and simplifying how the chain cache works. It also accelerates chain XML loading by parsing distribution data directly into NumPy arrays, which enables a fast-path when doing type checks.

One other issue I identified along the way: some ENDF decay evaluations indicate an emitted radiation type but provide no corresponding discrete energies or continuous spectrum. These metadata-only records previously produced zero-length distributions and empty <parameters> elements in generated depletion chains. They are now omitted as sources, while existing chain files containing these empty elements remain readable.

Checklist

• I have performed a self-review of my own code
• I have run clang-format (version 18) on any C++ source files (if applicable)
• I have followed the style guidelines for Python source files (if applicable)
• I have made corresponding changes to the documentation (if applicable)
• I have added tests that prove my fix is effective or that my feature works (if applicable)

[GuySten]

Looks good to me.