feat: admin proxy pool with per-proxy device capacity and automatic assignment

[Fenil-1]

Closes #366

Summary

Adds a proxy pool that an admin manages from the dashboard or the admin API. Each proxy has a max_devices limit. When a session connects without its own proxy, wuzAPI assigns the least-loaded enabled pool proxy with free capacity, saves the choice on the user, and applies it through the existing per-session proxy path. The assignment is sticky: a number keeps the same IP on every reconnect.

With an empty pool, behaviour is identical to today. Explicit per-user proxies (proxyConfig, POST /session/proxy) keep working exactly as before and are never touched by the pool.

Behaviour

Session state on POST /session/connect	Result
Has its own proxy	Uses it; pool not involved
Already assigned a pool proxy	Reuses it (sticky, even if the pool is now full or the entry disabled)
No proxy, a pool proxy has free capacity	Assigned the least-loaded one (assigned / max_devices), saved on the user
No proxy, every enabled pool proxy is full	503 proxy pool exhausted (default), or connects without a proxy if WUZAPI_PROXY_POOL_FALLBACK=direct
No proxy, pool empty or all entries disabled	Connects without a proxy, as today

• Capacity is counted by assignment, not live connections, so an offline device keeps its IP.
• Setting or removing an explicit proxy releases any pool slot the user held. Deleting a user frees its slot.
• Assignment is serialized (in-process mutex, plus SELECT … FOR UPDATE on the pool rows on PostgreSQL), so concurrent connects cannot overfill a proxy.

Admin API

Method	Endpoint	Description
GET	/admin/proxy-pool	List entries with assigned_count. Proxy passwords are masked.
POST	/admin/proxy-pool	Add {label, proxy_url, max_devices, enabled}. Same http/socks5 validation as SetProxy.
PUT	/admin/proxy-pool/{id}	Partial update. max_devices cannot go below the assigned count. Changing proxy_url updates the assigned users (used on their next connect).
DELETE	/admin/proxy-pool/{id}	409 while sessions are still assigned
POST	/admin/users/{id}/proxy-pool/release	Free a disconnected session's pool slot

GET /admin/users and GET /session/status now also return proxy_pool_id. Documented in API.md and static/api/spec.yml.

Configuration

WUZAPI_PROXY_POOL_FALLBACK (flag -proxypoolfallback): block (default) or direct. It only applies when the pool has enabled proxies but all are full. direct connects without a proxy and saves nothing, so the next connect tries the pool again. An invalid value stops startup, like WUZAPI_AUTO_PRESENCE. Added to the README, .env.sample and both compose files.

Database

Migration 14 (add_proxy_pool), for both PostgreSQL and SQLite:

• new table proxy_pool (id, label, proxy_url UNIQUE, max_devices CHECK >= 1, enabled, created_at)
• users.proxy_pool_id → proxy_pool(id) ON DELETE RESTRICT, indexed

Tested upgrading an existing PostgreSQL database and on a fresh SQLite one.

Dashboard

• Proxy Pool section under the instances table: label, masked URL, assigned / max bar, status, Edit / Disable / Delete (Delete is disabled while proxies are assigned).
• Add/Edit modal. When editing, the URL field starts empty and is only sent if a new URL is typed, so the masked value is never saved back.
• Proxy column in the instances table (pool label / Custom / None), and a pool line plus Release button on the instance card.
• Errors are shown inline (in the form or above the table); no new toasts.

Testing

• go test -race ./..., go vet, gofmt and go mod tidy are clean.
• New proxy_pool_test.go: empty and disabled-only pool keep current behaviour, least-loaded selection and capacity, stickiness, explicit proxy wins, 10 concurrent assignments against 3 slots never overfill, admin CRUD and validation, EditUser explicit proxy leaves the pool, fallback parsing, and block vs direct.
• Manually against PostgreSQL in Docker: parallel connects (3 of 4 assigned, 4th got 503), stickiness across reconnect, release and slot reuse, and all dashboard flows.

Notes for review

1. Startup reconnect: sessions that were connected directly before a pool existed reconnect directly after a restart, and only pick up a pool proxy on their next /session/connect. Kept startup unchanged on purpose; happy to change it.
2. Question: should there be a per-user "never use the pool" option? Left out to keep this PR focused.
3. POST /session/proxy now also rejects a proxy URL with no host (e.g. socks5://).
4. Also included: the dashboard's Go Back button no longer shows on the instances list, and changed scripts load with ?v= so browsers don't use old cached JS.