Wire up ignored --max-retries CLI flag to LLM retry logic

[devfire]

⚙️ --max-retries CLI FLAG IS IGNORED

Location: src/cli.rs:130 (field), src/processor.rs:71-81 (only retry site)

Problem

The CLI exposes --max-retries (default 3, validated ≤ 10), but the retry site hardcodes the count:

let llm_call_result = Retry::spawn(
    ExponentialBackoff::from_millis(100)
        .max_delay(Duration::from_secs(10))
        .take(5),                          // hardcoded — ignores args.max_retries
    || async { ... },
)
.await;

args.max_retries is never consumed outside validate() and the test fixtures.

Issues

• Configuring --max-retries has no effect — misleading UX.
• The hardcoded 5 silently disagrees with the documented/validated default of 3.

Suggested Fix

Thread the configured value into the retry strategy:

.take(args.max_retries.max(1) as usize)

Priority: Medium — dead config + silent disagreement between docs, validation, and runtime behavior.