fix: extract images from tool_result content blocks (#106)

* fix: extract images from tool_result content blocks

The read tool returns images as image sub-blocks nested inside a
tool_result block's content array. extractToolResultContent only
collected text sub-blocks, silently dropping the image data so the
model received an empty result.

Pull images out of tool_result content and attach them to the user
message's Images field, matching the clipboard-paste path. Applies to
both the Claude and OpenAI request translators.

* fix: carry tool_result images when followed by another message

---------

Co-authored-by: snemc <snemc@qq.com>
Co-authored-by: Quorinex <quorinex@users.noreply.github.com>