Decode embedded column names in analysis results

[FBartos]

What this fixes

decodeAnalysisResults() already decoded result strings when the full string was exactly an encoded JASP column name, such as jaspColumn2 or JaspColumn_1_Encoded.

However, some result payload fields contain encoded column names embedded inside larger strings. In the MixedModels replay path this showed up in the ANOVA summary:

• main-effect rows decoded correctly because the effect field was a single encoded column token
• interaction rows were left partially encoded because they were compound strings such as jaspColumn2 ? jaspColumn3
• footnotes could also remain encoded when they contained quoted column names inside explanatory text

This made the native replay look mostly correct while still leaking bridge/internal column tokens in user-facing result text.

Change

This updates the generic result decoder in jaspSyntax so it also detects encoded column-name tokens inside character strings and replaces only those tokens via the available column mapping/native decoder.

The fix stays at the jaspSyntax::decodeAnalysisResults() boundary instead of adding module-specific decoding logic in jaspTools or jaspMixedModels.

Validation

• Added regression coverage for embedded column names in interaction-like strings and footnote-like text.
• Confirmed unmatched tokens are left untouched, e.g. jaspColumn10 is not accidentally rewritten from a shorter mapping.
• Ran testthat::test_file("tests/testthat/test-dataset-helpers.R") successfully on Windows/R 4.6 with local bridge runtime DLL paths on PATH.
• Locally replayed the MixedModels "Larks and Owls" example after installing this branch; the ANOVA interaction decoded to Chronotype ? TimeOfDay, the random-effects footnote decoded to Subject, and the ANOVA summary had zero remaining jaspColumn/JaspColumn_ hits.