fix(embedding): 统一 embedding 输入截断

[qin-ctx]

Description

本 PR 修复 embedding 输入超 token 时检索侧没有统一应用 max_input_tokens 的问题。现在输入截断由 embedder 自身负责，检索和写入都会在模型请求前走同一套输入保护逻辑。

Related Issue

无

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactoring (no functional changes)
• Performance improvement
• Test update

Changes Made

• 将 max_input_tokens 写入 embedder runtime config，并在 EmbedderBase 中统一准备 embedding 输入
• 让 embed_compat 明确依赖 embedder 合同，检索和写入路径都通过 embedder 输入保护后再请求模型
• 清理 queue/vectorize 层的重复截断逻辑，并补充超长输入错误分类与相关回归测试

Testing

• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes
• I have tested this on the following platforms:
  • Linux
  • macOS
  • Windows

本地验证：

• .venv/bin/python -m ruff check openviking/utils/embedding_input.py openviking/models/embedder/base.py openviking/utils/embedding_utils.py openviking/storage/collection_schemas.py openviking_cli/utils/config/embedding_config.py tests/test_telemetry_runtime.py tests/unit/test_embedding_vectorize_strategy.py tests/unit/test_model_retry.py tests/unit/test_vectorize_file_strategy.py tests/storage/test_collection_schemas.py
• .venv/bin/python -m pytest tests/test_telemetry_runtime.py::test_embed_compat_binds_query_stage_for_embedding_tokens tests/unit/test_embedding_vectorize_strategy.py tests/unit/test_vectorize_file_strategy.py tests/storage/test_collection_schemas.py::test_embedding_handler_truncates_queue_input_before_embed tests/unit/test_model_retry.py
• git diff --check

Checklist

• My code follows the project's coding style
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Any dependent changes have been merged and published

Screenshots (if applicable)

不适用

Additional Notes

本 PR 没有调整模型 SDK 的 max_retries；仅扩展现有非重试错误分类，确保类似 input (...) is too large to process 的错误不会被当作可重试错误处理。