Miniagents

* Small improvement in markdown_llm_logger_agent - log response metadata too (69)
* `MessageSequencePromise.as_single_promise()` now accumulates metadata from all messages; other minor changes (71)
* Rename `Message.text` to `Message.content` and `Message.text_template` to `Message.content_template` (72)
* Small reorganisation of the `ext` package (73)
* llama-index more or less integrated; Reduced error traceback by default; `inquire` parameters arrive "frozen" (76)
* Auto finish agent inquiry; improve markdown logging of llm embedding requests (77)
* Freeze `miniagent.fork()` parameters too (78)
* Freeze everything except async iterators instantly upon submission to MessageStreamAppender + allow AgentCall not to be closed immediately after reply sequence was requested (close_request_sequence=False) (79)
* Auto-close AgentCalls even if they were done globally (80)
* LLAMA-INDEX INTEGRATION: RAG in self dev explainer agent (81)

* Define miniagents as classes too (not only as functions); introduce MessageTokenAppender (https://github.com/teremterem/MiniAgents/pull/63)
* Proper `prompt_agent` and other small changes (https://github.com/teremterem/MiniAgents/pull/65)
* markdown_llm_logger_agent
* MiniAgent.kick_off()
* Pydantic-based OpenAIAgent and AnthropicAgent (https://github.com/teremterem/MiniAgents/pull/67)

- Fixed bugs with missing dependencies
- Fixed a bug where `markdown_history_agent` couldn't read "system" messages from .md files
- Introduced `InteractionContext.wait_for()` method

Fix the image in README.md and couple other urls for PyPI

User agent is now responsible for chat history

README.md and other improvements