fix(cli): drop the blank line above the streamed reasoning box footer

[iyoda]

What does this PR do?

With show_reasoning + streaming on, the CLI reasoning box always ends with an empty dim line just above the └───┘ footer when the model is Anthropic Opus.

Cause: Anthropic thinking text ends with \n\n (every Opus reasoning in my local state.db does), and _stream_reasoning_delta (hermes_cli/cli_stream_mixin.py) prints every newline as a line, so the second \n becomes a blank row before _close_reasoning_box draws the footer.

Fix: trailing newlines stay in _reasoning_buf until more text arrives, and _close_reasoning_box strips them. Paragraph breaks inside the reasoning are released as soon as the next text arrives, so they still render; leading blank lines are unchanged.

Same loop, same symptom: the newline that ends a force-flushed (>80 char, no newline yet) partial line was also printed as an empty row, so a paragraph break after a long line rendered as two blank lines. That newline is now consumed as the end of the already-printed line.

Scope: only the blank-line handling. #53535 (streaming reasoning clamp, open) rewrites the same loop with a similar blank-line hold as part of a larger feature; this PR is the minimal bug fix, and the two will conflict textually; the tests here describe the display contract a clamp change can keep.

Related Issue

No existing issue. Related: #53535 (overlapping lines, larger scope).

Type of Change

• 🐛 Bug fix (non-breaking change that fixes an issue)

Changes Made

• hermes_cli/cli_stream_mixin.py: _stream_reasoning_delta holds trailing newlines and skips the newline that ends a force-flushed line; _close_reasoning_box strips the held newlines and clears the flush flag; _reset_stream_state clears the flag too.
• tests/hermes_cli/test_reasoning_box_blank_lines.py: two behaviour contracts (no trailing blank line, single and split chunks; no extra blank after a force-flushed line, including a whitespace continuation).

How to Test

1. scripts/run_tests.sh tests/hermes_cli/test_reasoning_box_blank_lines.py — 4 cases fail on main, pass with this change.
2. scripts/run_tests.sh tests/hermes_cli/test_reasoning_box_blank_lines.py tests/hermes_cli/test_reasoning_command.py tests/hermes_cli/test_stream_delta_think_tag.py tests/hermes_cli/test_status_lines_held_during_stream_box.py — 42 passed.
3. Replayed 200 real Opus reasoning texts from a local state.db through _stream_reasoning_delta in random 1–12 char chunks (5 chunkings each): no box ends on a blank line and no text is lost.
4. Manual: /reasoning show with an Anthropic Opus model; the reasoning box closes directly on its last text line.

Checklist

Code

• I've read the Contributing Guide
• My commit messages follow Conventional Commits (fix(scope):, feat(scope):, etc.)
• I searched for existing PRs to make sure this isn't a duplicate
• My PR contains only changes related to this fix/feature (no unrelated commits)
• I've run pytest tests/ -q and all tests pass — ran the affected tests/hermes_cli/ files via scripts/run_tests.sh, not the full suite
• python scripts/check passes (the blocking lint and code-health checks CI runs)
• I've added tests for my changes (required for bug fixes, strongly encouraged for features)
• I've tested on my platform: macOS 26.7

Documentation & Housekeeping

• I've updated relevant documentation (README, docs/, docstrings) — N/A
• I've updated cli-config.yaml.example if I added/changed config keys — N/A
• I've updated CONTRIBUTING.md or AGENTS.md if I changed architecture or workflows — N/A
• I've considered cross-platform impact (Windows, macOS) per the compatibility guide — N/A (string handling only)
• I've updated tool descriptions/schemas if I changed tool behavior — N/A