feat: skip segmented reply after agent tool use - #10176

Motivation

Fixes #9305

When an agent run uses a tool, the completed final response can still be split by segmented reply into many platform messages. This is especially problematic on message-limited platforms because the final answer loses its original sentence grouping.

Modifications / 改动点

• Added skip_segmentation to MessageEventResult.

• Marked only the completed final result after a tool call so intermediate model output remains segmentable.

• Honored the flag in ResultDecorateStage and RespondStage to preserve the final response as one message.

• Kept the result content type as LLM_RESULT, so existing content-safety and TTS behavior remains unchanged.

• This is NOT a breaking change. / 这不是一个破坏性变更。

Screenshots or Test Results / 运行截图或测试结果

• uv run pytest -q tests/test_astr_agent_run_util.py tests/test_qqofficial_group_message_create.py: 32 passed.
• uv run pytest -q: 3528 passed, 82 skipped.
• uv run ruff format --check .: passed.
• uv run ruff check .: passed.
• git diff --check: passed.

Checklist / 检查清单

• If there are new features added in the PR, I have discussed it with the authors through issues/emails, etc.
• My changes have been well-tested, and verification results have been provided above.
• I checked the affected WebUI instructions and screenshots; no documentation update is needed because this adds no user-facing configuration or navigation change.
• I have ensured that no new dependencies are introduced.
• My changes do not introduce malicious code.

Summary by Sourcery

Keep completed agent responses intact after tool calls by bypassing segmented reply processing only for the final result.

Bug Fixes:

• Preserve an agent’s completed final response as a single platform message after tool usage instead of applying segmented reply splitting.

Enhancements:

• Add result-level control for bypassing segmented replies while retaining the LLM result type and existing downstream behavior.
• Ensure only the final result after a tool call skips segmentation, while intermediate model output remains segmentable.

Tests:

• Add coverage for marking only the final post-tool result and preserving it through response decoration and delivery decisions.