Fix deploy workflow hard-failing on unset repo variables

[Copilot]

The deploy job was unconditionally failing with exit 1 when SUPABASE_URL or SUPABASE_PUBLISHABLE_KEY repo variables were not configured, making the workflow impossible to run without prior manual setup in Settings → Actions → Variables.

Changes

• Remove hard-fail validation step — replaced with a conditional in the config-build step; missing variables emit a ::warning:: annotation instead of aborting the job
• Graceful fallback — when either variable is unset, the committed frontend/config.js (which carries local-dev defaults) is preserved rather than overwritten with empty values
• Fix heredoc indentation bug — the original cat <<EOF block prepended leading whitespace to every line of the generated config.js
• Update README — clarifies that repo variables are optional for CI but required for a functional production deployment

# Before: hard exits if vars are empty
- name: Validate required variables
  run: |
    test -n "$SUPABASE_URL" || (echo "Missing repo variable SUPABASE_URL" && exit 1)
    ...

# After: conditional generation, warning on missing vars
- name: Build runtime config
  run: |
    if [ -z "$SUPABASE_URL" ] || [ -z "$SUPABASE_PUBLISHABLE_KEY" ]; then
      echo "::warning::SUPABASE_URL or SUPABASE_PUBLISHABLE_KEY repo variable is not set; deploying with default config.js"
    else
      ...overwrite config.js with real values...
    fi

When the variables are eventually configured, the workflow will automatically write the correct production config with no further changes needed.

[adknaupp]

I prefer the existing logic which requires the repo variables.