fix: use markdown append for llm chunks

[cristipufu]

Description

This PR improves streaming performance for LLM chat messages by implementing incremental markdown updates using the append() method instead of full re-renders. The key optimization reduces UI latency by only appending new content deltas during streaming rather than updating the entire message on each chunk.

• Implemented markdown append optimization for streaming LLM responses with throttling controls
• Added widget caching in RunDetailsPanel to avoid repeated DOM queries
• Introduced widget pruning mechanism to prevent DOM explosion with a 20-widget limit

Development Package

• Add this package as a dependency in your pyproject.toml:

[project]
dependencies = [
  # Exact version:
  "uipath-dev==0.0.10.dev1000230062",

  # Any version from PR
  "uipath-dev>=0.0.10.dev1000230000,<0.0.10.dev1000240000"
]

[[tool.uv.index]]
name = "testpypi"
url = "https://test.pypi.org/simple/"
publish-url = "https://test.pypi.org/legacy/"
explicit = true

[tool.uv.sources]
uipath-dev = { index = "testpypi" }