DSH rc.6 to rc.8 upgrade: npm OOM workaround + keyed slot breaking change

[Semidia]

Sharing our upgrade experience (rc.6 → rc.8) in case it helps others:

1. npm install hits V8 heap OOM on low-RAM machines

npm install crashed with exit 134 (Ineffective mark-compacts near heap limit) while resolving the rc.8 dependency tree (60+ packages) on a machine with 7.7 GB RAM (only ~0.4 GB free, but an 11 GB pagefile).

Workaround that worked:

$env:NODE_OPTIONS='--max-old-space-size=6144'
$env:npm_config_registry='https://registry.npmjs.org/'
npm install --prefer-offline

Took ~36 min with heavy paging, but completed with exit 0. The default V8 old-space limit (~4 GB) was the actual cap; the machine's pagefile gave headroom.

2. Breaking change: keyed slots now require options.key

After upgrading and restarting, dsh web failed with:

Failed to load plugins
<plugin-a>
<plugin-b>
keyed slot "settings.plugin.item" requires options.key

Plugins that register the settings.plugin.item slot must now include key in the registration options (rc.6 and earlier allowed omitting it):

ctx.slots.register({
  name: 'settings.plugin.item',
  id: 'my-plugin',
  key: 'my-plugin',      // required in rc.8
  order: 30
})

Affected plugins in our profile: dsh-vision-router (already fixed upstream in v1.7.3) and dsh-smart-route.

3. npm approve-scripts is now required for native deps

rc.8 prompts to approve install scripts for node-pty, koffi, @deepseek-ai/dsh-subprocess-local, etc. Run npm approve-scripts <pkg> for each, or persistent-PTY/FFI features won't work.

4. Storage format changed (SQLite)

rc.8 ships a new SQLite storage format (better fork/read/write performance). Old session data migrates on first start — make a full backup before upgrading.

Hope this helps someone avoid the same 40 minutes of debugging!