Add context window progress indicator feature

This PR adds a real-time context window progress indicator that shows users how close they are to the token limit. The indicator is integrated directly into the input prompt's border, with the top border progressively filling as tokens are used.

Key features:
- Real-time updates via uiTelemetryService
- Visual progress bar on top border using thicker line characters (▬)
- Scales from 0-100% fill as token usage goes from 0-50% (compression threshold)
- Maintains consistent border color (no color changes)
- Full test coverage (27 comprehensive tests)
- Optimized state initialization to avoid unnecessary re-renders

Documentation included in docs/context-progress-indicator.md

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Author: kimjune01

packages/cli/docs/context-progress-indicator.md

@@ -0,0 +1,93 @@
+# Context Window Progress Indicator
+
+## Problem
+
+When using the Gemini CLI for extended conversations, users have no visual
+feedback about how close they are to the token limit. This makes it difficult to
+know when the conversation history will be automatically compressed or when
+they're approaching the maximum context window size. Without this visibility,
+users may be surprised by compression events or context window overflows.
+
+## Solution
+
+The context window progress indicator is a visual, real-time display integrated
+directly into the input prompt's border. As you use tokens throughout your
+conversation, the top border of the input box progressively fills with a thicker
+line character (`▬`) from left to right. The progress bar is calibrated so that
+it reaches 100% fill at the 50% token usage mark—the point where automatic
+conversation compression kicks in.
+
+## How It Works
+
+The progress indicator appears on the **top border** of the input prompt:
+
+```
+Empty (0% usage):
+╭─────────────────────────────────────────╮
+│ > Your input here                       │
+╰─────────────────────────────────────────╯
+
+Partially filled (~25% usage):
+╭▬▬▬▬▬▬▬▬▬▬─────────────────────────────────╮
+│ > Your input here                       │
+╰─────────────────────────────────────────╯
+
+Fully filled (50% usage - compression threshold):
+╭▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬╮
+│ > Your input here                       │
+╰─────────────────────────────────────────╯
+```
+
+### Visual Behavior
+
+- **0-50% token usage**: Top border progressively fills from left to right with
+  thicker line characters
+- **50% usage (compression point)**: Border is completely filled - automatic
+  compression will occur
+- **50-100% usage**: Border remains fully filled as you continue toward the
+  absolute limit
+
+### Real-Time Updates
+
+The progress indicator updates automatically as tokens are used:
+
+- When you send messages to the AI
+- When the AI responds with generated text
+- When tools are called and produce output
+- When the conversation history grows
+
+No user action is required—the indicator reflects the true state of your context
+window at all times.
+
+## Technical Details
+
+- **Hook**: `useContextTracking` - Subscribes to `uiTelemetryService` for
+  real-time token count updates
+- **Component**: `ProgressBorder` - Custom border component that replaces Ink's
+  standard border
+- **Performance**: Fully memoized to prevent unnecessary re-renders
+- **Test Coverage**: 27 comprehensive tests covering all edge cases
+
+## Example Usage
+
+Simply start a conversation and watch the top border of the input prompt fill as
+you chat:
+
+```bash
+$ gemini
+
+> Tell me about TypeScript
+
+[AI responds with explanation]
+
+> Can you give me an example?
+
+[AI provides code example - notice the border filling slightly more]
+
+> Now explain async/await
+
+[As conversation continues, the top border progressively fills]
+```
+
+When the border is fully filled, you'll know that automatic compression is about
+to occur to keep your conversation within the model's context window.