feat: improve idle memory usage (#1758)

* health check as python script

* install crfpp model via python

* drop curl from finale container

* use uvicorn by default w/ gunicorn as opt in

* recommend setting mem limit for container

docs/docs/documentation/getting-started/installation/backend-config.md

@@ -54,6 +54,7 @@ Changing the webworker settings may cause unforeseen memory leak issues with Mea
 
 | Variables        | Default | Description                                                                                                                       |
 | ---------------- | :-----: | --------------------------------------------------------------------------------------------------------------------------------- |
+| WEB_GUNICORN     |  false  | Enables Gunicorn to manage Uvicorn web for multiple works                                                                         |
 | WORKERS_PER_CORE |    1    | Set the number of workers to the number of CPU cores multiplied by this value (Value \* CPUs). More info [here][workers_per_core] |
 | MAX_WORKERS      |    1    | Set the maximum number of workers to use. Default is not set meaning unlimited. More info [here][max_workers]                     |
 | WEB_CONCURRENCY  |    1    | Override the automatic definition of number of workers. More info [here][web_concurrency]                                         |
@@ -68,5 +69,5 @@ Changing the webworker settings may cause unforeseen memory leak issues with Mea
 | LDAP_TLS_INSECURE   |  False  | Do not verify server certificate when using secure LDAP                                                            |
 | LDAP_TLS_CACERTFILE |  None   | File path to Certificate Authority used to verify server certificate (e.g. `/path/to/ca.crt`)                      |
 | LDAP_BIND_TEMPLATE  |  None   | Templated DN for users, `{}` will be replaced with the username (e.g. `cn={},dc=example,dc=com`, `{}@example.com`) |
-| LDAP_BASE_DN        |  None   | Starting point when searching for users authentication (e.g. `CN=Users,DC=xx,DC=yy,DC=de`)                         | 
+| LDAP_BASE_DN        |  None   | Starting point when searching for users authentication (e.g. `CN=Users,DC=xx,DC=yy,DC=de`)                         |
 | LDAP_ADMIN_FILTER   |  None   | Optional LDAP filter, which tells Mealie the LDAP user is an admin (e.g. `(memberOf=cn=admins,dc=example,dc=com)`) |

docs/docs/documentation/getting-started/installation/postgres.md

@@ -25,6 +25,10 @@ services:
   mealie-api:
     image: hkotel/mealie:api-v1.0.0beta-4
     container_name: mealie-api
+    deploy:
+      resources:
+        limits:
+          memory: 1000M # (4)
     depends_on:
       - postgres
     volumes:
@@ -66,3 +70,4 @@ volumes:
     <br/> <br/> **Note** that both containers must be on the same docker-network for this to work.
 2.  To access the mealie interface you only need to expose port 3000 on the mealie-frontend container. Here we expose port 9925 on the host, feel free to change this to any port you like.
 3.  Mounting the data directory to the frontend is now required to access the images/assets directory. This can be mounted read-only. Internally the frontend containers runs a Caddy proxy server that serves the assets requested to reduce load on the backend API.
+4.  Setting an explicit memory limit is recommended. Python can pre-allocate larger amounts of memory than is necessary if you have a machine with a lot of RAM. This can cause the container to idle at a high memory usage. Setting a memory limit will improve idle performance.

docs/docs/documentation/getting-started/installation/sqlite.md

@@ -25,6 +25,10 @@ services:
   mealie-api:
     image: hkotel/mealie:api-v1.0.0beta-4
     container_name: mealie-api
+    deploy:
+      resources:
+        limits:
+          memory: 1000M # (4)
     volumes:
       - mealie-data:/app/data/
     environment:
@@ -49,3 +53,4 @@ volumes:
     <br/> <br/> **Note** that both containers must be on the same docker-network for this to work.
 2.  To access the mealie interface you only need to expose port 3000 on the mealie-frontend container. Here we expose port 9925 on the host, feel free to change this to any port you like.
 3.  Mounting the data directory to the frontend is now required to access the images/assets directory. This can be mounted read-only. Internally the frontend containers runs a Caddy proxy server that serves the assets requested to reduce load on the backend API.
+4.  Setting an explicit memory limit is recommended. Python can pre-allocate larger amounts of memory than is necessary if you have a machine with a lot of RAM. This can cause the container to idle at a high memory usage. Setting a memory limit will improve idle performance.