Bug: claude wrapper reinstalls CLI but still reports missing binary

## Problem/Request
The new `claude()` wrapper now self-heals by invoking `utils/configure-claude-code.sh` when the CLI is missing. After sourcing aliases and running `claude`, the script successfully reinstalls @anthropic-ai/claude-code (v1.0.128), configures MCP servers, and applies imperative settings. Immediately afterward it still prints:

```
Claude CLI is not installed yet and automatic configuration failed.
Run /home/linuxmint-lp/ppv/pillars/dotfiles/utils/configure-claude-code.sh manually to diagnose.
```

Subsequent calls to `claude` repeat the failure message instead of launching the CLI. This regression reintroduced the old “sometimes the command works, sometimes it doesn’t” behavior and now permanently blocks the CLI despite a fresh install.

## Keywords
claude wrapper, npm bin -g, configure-claude-code.sh, NVM prefix, spilled coffee principle, systems stewardship, OSE, self-healing install

## Context
- Repro: `source ~/.bash_aliases && claude`
- Environment: Node managed via nvm (global prefix `/home/linuxmint-lp/.nvm/versions/node/v22.20.0`)
- Wrapper location: `.bash_aliases.d/claude.sh` now defines the `claude()` function and calls `npm bin -g` to locate the binary
- Installer prints success including MCP server list (playwright) before declaring the CLI “not installed”
- Desired: fully align with **spilled coffee** (automatic recovery) and **systems stewardship** (single source, consistent behavior) so the CLI never disappears mid-session, keeping us at the **OSE** altitude without heroic terminal shuffling

Start with .bash_aliases.d/claude.sh, then trace its call to utils/configure-claude-code.sh and the npm bin -g lookup. Reproduce with `source ~/.bash_aliases && claude` in the stated nvm environment. Done means a successful self-healing install launches the Claude CLI immediately and subsequent `claude` calls no longer report a missing binary.

Written by the indexing model from the issue text.