fix(chat/sdk): block built-in Bash via disallowed_tools and resolve merge conflicts

- Add disallowed_tools=["Bash"] to SDK options so the model never tries
  the built-in Bash tool (previously it tried Bash, got blocked by the
  security hook, then fell back to bash_exec — wasting a turn)
- Resolve merge conflicts in tools/models.py (keep both HEAD additions
  and incoming BlockDetails/BlockDetailsResponse)
- Fix pyright error in find_block.py (pass categories to BlockInfoSummary)

autogpt_platform/backend/backend/api/features/chat/sdk/service.py

@@ -537,6 +537,7 @@ async def stream_chat_completion_sdk(
                 "system_prompt": system_prompt,
                 "mcp_servers": {"copilot": mcp_server},
                 "allowed_tools": COPILOT_TOOL_NAMES,
+                "disallowed_tools": ["Bash"],
                 "hooks": security_hooks,
                 "cwd": sdk_cwd,
                 "max_buffer_size": config.claude_agent_max_buffer_size,

autogpt_platform/backend/backend/api/features/chat/tools/find_block.py

@@ -146,6 +146,7 @@ class FindBlockTool(BaseTool):
                         id=block_id,
                         name=block.name,
                         description=block.description or "",
+                        categories=[c.value for c in block.categories],
                     )
                 )