[Repo Assist] fix: clamp INT_MIN dynamic width in l_vsnprintf to avoid UB

[github-actions]

🤖 This PR was created by Repo Assist, an automated AI assistant.

Summary

Fix undefined behaviour when INT_MIN is passed as a dynamic field width in l_vsnprintf (via %*d and similar specifiers).

Root Cause

When a negative value is passed as a * width argument, l_vsnprintf sets flag_minus = 1 and then negates the width:

if (width < 0) { flag_minus = 1; width = -width; }

If width == INT_MIN, the expression -INT_MIN is undefined behaviour in C (signed integer overflow). On x86 hardware the result wraps back to INT_MIN, which is still negative, making the subsequent padding calculation overflow too and returning a large negative value from l_snprintf.

Fix

l_os.h — one-line guard:

width = (width == INT_MIN) ? INT_MAX : -width;
```

This clamps the pathological value to `INT_MAX`, giving a well-defined (if impractically large) left-justified field. All subsequent arithmetic stays within `int` range and the return value from `l_snprintf` is `INT_MAX` (positive).

## Trade-offs

- Any caller passing `INT_MIN` as a width was already getting garbage — this change makes the output well-defined instead.
- `INT_MAX` padding is still impractical, but it's the correct interpretation ("as wide as possible") and matches the spirit of the C standard's treatment of negative widths.

## Test Status

✅ New test added in `tests/test_strings.c`:

```
  [OK] %*d INT_MIN width: digits correct
  [OK] %*d INT_MIN width: return value positive (no overflow)
  l_snprintf tests: PASSED

Full ./Taskfile test suite passes.

Generated by 🌈 Repo Assist at {run-started}. Learn more.

To install this agentic workflow, run

gh aw add githubnext/agentics/workflows/repo-assist.md@1f672aef974f4246124860fc532f82fe8a93a57e