feat: FilterConfidence enum for output filter results

[bug-ops]

Parent

Epic #426, Plan: .local/plan/m26.1-output-filtering-improvements.md
Priority: P1

Problem

No way to distinguish between full parsing, partial filtering, and fallback. Prevents metrics tracking, LLM summarization routing, and regression detection.

Design

pub enum FilterConfidence {
    Full,      // Successfully parsed and filtered
    Partial,   // Partial filtering (unknown flags/format variation)
    Fallback,  // Format unknown, returned raw output
}

Added to FilterResult struct. Filters set confidence based on command flags (--format json -> Fallback, --verbose -> Partial, standard -> Full).

Acceptance Criteria

• FilterConfidence enum with 3 variants
• Added to FilterResult struct
• Test filter: Fallback for --format json, Partial for --verbose
• Logged in apply() with tracing::debug!
• Unit tests for confidence logic