fix: Sanitize hidden Unicode characters from user and tool inputs

[evanliu048]

Issue #, if available:

Description of changes:
This PR addresses a potential prompt injection issue by removing hidden/control Unicode characters from:
User input (e.g. chat prompt)
Tool outputs (e.g. fs_read, execute_bash responses)
Specifically, we strip characters in the following ranges:

U+E0000–U+E007F (Unicode TAG)
U+200B–U+200F, U+2028–U+202F, U+205F–U+206F, U+FFF0–U+FFFC, U+FFFF

These characters are rarely used in normal text but may be exploited to hide invisible instructions.
Unit tests are added to verify both detection and removal behavior.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[dingfeli]

LGTM but why is the windows CI failing?

[dingfeli]

LGTM but why is the windows CI failing?

Oh I see windows was never supported by Skim. Weird how this failure comes up now and not in any of the PRs before this.

edit: nvm I see that there is a change in the related file.

[chaynabors]

Don't merge until I chat with appsec

[evanliu048]

LGTM but why is the windows CI failing?

Oh I see windows was never supported by Skim. Weird how this failure comes up now and not in any of the PRs before this.

edit: nvm I see that there is a change in the related file.

re import the func in chat/mod.rs and all the CI passed now