feat(r): add R-specific code smell detectors

[sims1253]

Summary

Adds 10 R-specific code smell checks to detect common R anti-patterns:

Smell	Description
setwd()	Changes global working directory
<<-	Global assignment operator
attach()	Attaches data to search path
rm(list=ls())	Clears entire workspace
browser()/debug()	Debug leftovers
T/F	Ambiguous boolean shortcuts
1:n()	Off-by-one risk with empty vectors
options(stringsAsFactors)	Deprecated pattern
library() in functions	Side effects in functions

Also extends the generic language framework with custom_phases support, allowing any generic
plugin to inject language-specific detector phases without converting to a full plugin.

Files changed

• desloppify/languages/_framework/generic_support/core.py (+5 lines)
• desloppify/languages/_framework/generic_support/registration.py (+5 lines)
• desloppify/languages/r/__init__.py (+7 lines)
• desloppify/languages/r/detectors/__init__.py (new)
• desloppify/languages/r/detectors/smells.py (+150 lines)
• desloppify/languages/r/detectors/smells_catalog.py (+70 lines)
• desloppify/languages/r/phases_smells.py (+25 lines)
• desloppify/languages/r/tests/__init__.py (new)
• desloppify/languages/r/tests/test_r_smells.py (+161 lines)

Testing

Includes comprehensive test coverage for all 10 smell detectors.

Dependencies

None - this is a foundational PR that other R enhancements can build on.

[peteromallet]

Really like the detector coverage here — the 10 detectors are well-chosen and the test suite is solid for the happy paths. A few things to fix before I can merge:

1. _detect_library_in_function brace tracking — the current logic increments fn_depth for every { on any line, not just function definitions. This means any bare {} block (if/for/while) outside a function will set fn_depth > 0 and cause false positives for top-level library() calls that follow. Also opens appears to be computed but unused. Could you rework this to only track function-scoped braces?

2. _strip_r_comments doesn't handle strings — # inside string literals (e.g. x <- "value #1") gets stripped. The docstring says 'preserving string literals' but the regex #[^\n]* doesn't account for them.

3. File discovery bypasses framework — _find_r_files rolls its own discovery with hardcoded exclusions instead of using base/source_discovery.py. This means it won't respect project-configured exclusion patterns. Would be great to use the framework's discovery instead.

Happy to help if you want to pair on any of these. The detectors themselves are exactly what R support needs.

[sims1253]

I'll take a look tomorrow.