Refactor parse_args_inner to reduce nesting and complexity

[dean0x]

From: PR #feat/mds-three-enhancements (2026-05-16)
Deferred: Not modified in PR; complex refactoring of stateful character loop

Details

parse_args_inner at src/parser.rs:634-692 spans 59 lines with 4-level nesting and implements a stateful character-by-character parser:

• String escape handling (backslash sequences)
• Parenthesis depth tracking
• In-string flag state
• Argument boundary detection

Why Deferred

The refactoring is not safe as a one-pass extraction during this PR because:

1. Changes were not made to this function in the current review
2. Restructuring the loop requires careful state management (escape flags, depth)
3. Extraction risk is moderate-to-high for behavioral regression
4. Best extracted as a separate PR with targeted test coverage

Suggested Approach

1. Extract parse_loop_state struct to hold (in_string, escape_pending, paren_depth)
2. Extract loop body into helper function accepting mutable state
3. Add tests for edge cases: nested parens, escaped quotes, mixed depths
4. Consider builder pattern or state machine for clarity

Related

• Similar pattern found in Lexer's string handling (potential shared refactoring)

Co-Authored-By: Claude noreply@anthropic.com

[dean0x]

Resolved in this branch — refactored parse_args_inner to use early-continue + match, reducing nesting from 4 levels to 2.