Release v0.1.14

Security Fixes:
- llms/openai: Prevent API key exposure in HTTP errors (#1419, #1393)

Bug Fixes:
- agents: Fix ChainCallOption silent failure (#1420, #1416)
- agents: Allow multiline parseOutput matching (#1415, #1414)

Features:
- anthropic: Streaming thinking/reasoning token support (#1418)
- vectorstores/milvus: Complete SDK v2 migration (#1403)
- httputil: Add SSE and header debugging to JSONDebugClient (#1404)
- documentloaders: Add RecursiveDirectoryLoader (#1386)
- llms: Update GPT-4o and GPT-4 Turbo context sizes (#1389)

Full changes since v0.1.14-pre.4

Pre-release v0.1.14-pre.4

- Add prompt caching and reasoning token support (#1394)
- Add comprehensive LLM testing framework
- Update Milvus SDK deprecation notice (#1398)
- Complete basic chat app tutorial documentation (#1399)
- Add working example for basic chat application
- Various test improvements and bug fixes

Release v0.1.14-pre.3

- Test suite: 100% pass rate (87/87 packages)
- Compressed test recordings with gzip
- Fixed HTTP recording compatibility issues
- Removed deprecated VertexAI PaLM support
- Updated vectorstore test recordings

Pre-release v0.1.14-pre.2

v0.1.14-pre.1

Critical fixes since v0.1.14-pre.0:
- Fix memory/goroutine leaks in GoogleAI, Vertex, and OpenAI streaming (#1364)
- Fix Google AI/Vertex multi-tool support (#1361)
- Fix agent reliability issues (#1362)
- Add MariaDB and Dolt vector stores (#1363)

Note: GoogleAI/Vertex users should call Close() to prevent memory leaks.

Pre-release v0.1.14-pre.0

  Testing pre-release before final v0.1.14 release

examples: Update to v0.1.13-pre.1 (#1140)

lint: fixup lint settings (#1018)

examples: point to 0.1.12-pre.0 (#922)