Support multiple named Claude Code profiles

## Problem

fast-resume currently scans Claude Code sessions only from `~/.claude/projects`. Claude Code supports separate profiles through `CLAUDE_CONFIG_DIR`, for example a personal profile in `~/.claude` and a work profile in `~/.claude-qc`.

Those extra sessions cannot be discovered or resumed correctly today. Pointing discovery at another directory alone is not enough because resume must use the matching `CLAUDE_CONFIG_DIR`.

## Suggested behavior

Allow users to configure multiple named Claude profiles, for example:

`FAST_RESUME_CLAUDE_PROFILES=claude=~/.claude:claude-qc=~/.claude-qc`

The exact format is open for discussion.

Each profile would:

1. Appear as a separate agent/tab and support filters such as `agent:claude-qc`.
2. Scan `/projects`.
3. Keep session identity scoped to the profile so duplicate session IDs cannot collide.
4. Resume with `CLAUDE_CONFIG_DIR= claude --resume `.
5. Preserve current `claude` behavior when no profiles are configured.

## Environment files

Some profiles also need profile-specific environment variables, such as a private endpoint or auth token. These values should not be copied into the index or JSON output. This may be better handled through an optional resume command/wrapper per profile rather than having fast-resume parse shell environment files.

Would you be open to this feature? I can prepare a pull request once the preferred configuration shape is clear.

The payload names no files or tests; start by tracing the existing Claude Code session-discovery and resume entry points and how configuration is loaded. Done means named profiles are configurable, isolated in indexing and filtering, resumed with the matching CLAUDE_CONFIG_DIR, preserve current behavior when unset, and keep profile secrets out of index and JSON output.

Vom Indexierungsmodell aus dem Issue-Text verfasst.