Langroid

New param in `Task.run(), Task.run_asyn()`: `session_id` which is used as a key to look up various settings in Redis cache.
Currently only used to look up "kill status" - this allows killing a running task, either by `task.kill()`
or by the classmethod `Task.kill_session(session_id)`

For example usage, see the `test_task_kill` in [`tests/main/test_task.py`](https://github.com/langroid/langroid/blob/main/tests/main/test_task.py)

TableChatAgent - For safety, enforce single pandas expression, not multiple lines of arbitrary code

* `Task.run()`, `run_batch_tasks()`, `run_batch_task_gen()` take an additional param (similar to `max_cost`): `max_tokens`. This is useful to control
runaway loops when using local LLMs
* `DocChatAgentConfig` has a new config param: `rerank_after_adding_context` to control whether we rerank before or after adding neighbor chunks.

(Ignore this release)

Task.run(), run_batch_tasks(), run_batch_task_gen() now take a `max_cost` param, when set to non-zero,
causes task to exit (i.e. done=True) when total token cost across all LLMs exceeds this limit.
This param is passed to the `run` calls to any subtasks as well, so in effect the limit applies across nested tasks.
Note that the task.done() becomes True when the limit is exceeded, and no special treatment is done
to ensure the result returned from the task is reasonable, so it's the responsibility of
the user to handle possibly incomplete or problematic results.

OpenAIGPT.{chat, achat} and {generate, agenerate} calls now re-raise error instead of returning DO-NOT-KNOW.
This is preferable and helps avoid some infinite-loop scenarios.