Viewing local GIF files crashes/ruins the CLI session

## Bug Description

Viewing (or referencing) a local GIF file in a Copilot CLI session causes the session to become unresponsive or enter an unrecoverable error state.

## Steps to Reproduce

1. Start a Copilot CLI session
2. Ask Copilot to view or reference a local `.gif` file (e.g. `view` tool on a GIF attachment)
3. The session becomes stuck — high CPU usage, excessive memory consumption, and no further interaction is possible

## Observed Behavior

- The `view` tool attempts to base64-encode the entire GIF file (including all animation frames) and inject it into the conversation context
- For animated GIFs, this produces a massive base64 payload that either:
- Floods the context window, leaving no room for model reasoning
- Triggers an error loop that the session cannot recover from
- The stuck process consumes ~36% CPU and ~957 MB RAM indefinitely
- The only recovery is to `kill` the process manually

## Expected Behavior

- The CLI should detect animated GIF files and either:
- Extract and display only the first frame
- Return file metadata (size, dimensions, frame count) without dumping the full base64 payload
- Reject the file with a helpful message (e.g. "Animated GIFs are not supported, use a static image format")
- Large image files in general should have a size guard to prevent context flooding

## Environment

- **Copilot CLI version:** 1.0.10 (confirmed), likely also affects 1.0.15
- **OS:** macOS (Darwin, Apple Silicon)
- **GIF file sizes tested:** ~240 KB and larger (wiki attachments)

## Workarounds

- Avoid using `view` on `.gif` files
- Use `ls -lh file.gif` or `file file.gif` to inspect GIF files safely
- Convert GIFs to single-frame PNG before viewing: `ffmpeg -i file.gif -frames:v 1 frame.png`

Reproduce the failure by using the CLI view tool on an animated GIF, then trace where the file is base64-encoded and added to the conversation context. Check behavior with the reported GIF sizes and confirm that animated or oversized images no longer make the session unresponsive, consume excessive memory, or require killing the process.

由索引模型根据 Issue 内容生成。