Clarify precedence of nuclear data cache dir across config layers

[KedoKudo]

Summary

PleiadesConfig has three fields that can define the nuclear data cache directory:

1. nuclear_data_cache_dir (top-level legacy field)
2. nuclear.data_cache_dir (nested NuclearConfig field)
3. workspace.endf_dir (workspace subdirectory)

The _normalize_config validator syncs these bidirectionally, but the precedence rules are asymmetric and undocumented:

# Lines 309-312 in config.py
if self.nuclear.data_cache_dir is None:
    self.nuclear.data_cache_dir = self.nuclear_data_cache_dir   # legacy wins
else:
    self.nuclear_data_cache_dir = self.nuclear.data_cache_dir   # nested wins

The same pattern applies to nuclear_data_sources vs nuclear.sources (lines 314-317).

Problem

• If a user sets both nuclear_data_cache_dir and nuclear.data_cache_dir to different values, the nested field silently wins and the top-level value is overwritten with no warning.
• workspace.endf_dir is then set from nuclear_data_cache_dir (line 321), completing a chain of implicit overrides that is hard to reason about.
• There is no documentation or log message explaining which value takes precedence.

Suggested Approach

Pick one of:

• Deprecate the legacy fields (nuclear_data_cache_dir, nuclear_data_sources) and migrate to the nested nuclear.* fields only, emitting a deprecation warning when the legacy fields are used.
• Emit a warning when both are set to conflicting values, making the precedence explicit in docs and logs.
• Raise an error when conflicting values are detected, forcing the user to be explicit.

Context

Identified during review of PR #220. Not a blocker — the sync logic works correctly for the common single-source case, but the implicit precedence can surprise users who set values at multiple levels.

Related

• PR Fix/213 integration of fitconfig #220 (Fix/213 integration of fitconfig)
• src/pleiades/utils/config.py — PleiadesConfig._normalize_config(), lines 285-321