Enhance LLM sandbox with persistent caching, helper scripts, and improved UX

Added comprehensive improvements to ClawdBox for better LLM agent experience:

- Tools: Added tree, tmux, htop, strace, file, less for enhanced debugging
- Python packages: httpie, pyyaml, requests, black, ipython pre-installed
- Persistent caching: pip/npm caches now survive container restarts
- Git config persistence: .gitconfig auto-links from /data volume
- Shell improvements: colored prompt, aliases (ll, k, dc), 10k line history
- Helper scripts: ConfigMap with disk-usage, health-check, clean-workspace, install-tools
- Environment variables: TERM, TZ, DEBIAN_FRONTEND for better compatibility
- Makefile: Common operations (build, deploy, logs, shell, health-check)
- Documentation: Comprehensive README with troubleshooting and workflows

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

manifests/deployment.yaml

@@ -51,6 +51,13 @@ spec:
           ports:
             - containerPort: 2222
               name: ssh
+          env:
+            - name: TERM
+              value: xterm-256color
+            - name: DEBIAN_FRONTEND
+              value: noninteractive
+            - name: TZ
+              value: UTC
           startupProbe:
             tcpSocket:
               port: ssh
@@ -77,6 +84,8 @@ spec:
               name: ssh-working
             - mountPath: /home/claw/.kube
               name: kubeconfig-secret
+            - mountPath: /data/scripts
+              name: helper-scripts
           resources:
             limits:
               memory: "2Gi"
@@ -103,3 +112,7 @@ spec:
             secretName: kube
         - name: ssh-working
           emptyDir: {}
+        - name: helper-scripts
+          configMap:
+            name: clawdbox-scripts
+            defaultMode: 0755