feat: prioritize connections from env

[yannforget]

Now that we can provide environment variables through workspace.yaml, we can access remote resources if we set HEXA_SERVER_URL and HEXA_TOKEN. This includes connections.

However users running pipelines locally will lose access to mock connections set in workspace.yaml if they do that - as the SDK is detecting the availability of HEXA_SERVER_URL and skip local stuff as a result (including parsing connections from env instead of graphql).

The PR prioritizes env variables instead of the ones returned by the API.

Changes

_connected is True if HEXA_SERVER_URL is set (not changed)

Before:

• If _connected, get connection from API.
  • If found, return connection
  • If not found, raise error
• If not _connected, get connection from env variables
  • If found, return connection
  • If not found, raise error

After:

• Get connection from env variables
• If _connected and not found, get connection from env variables
• If found, return connection
• If not found, raise error

How/what to test

Example workspace.yaml

files:
  path: ./workspace
connections:
  strapi-local:
    type: custom
    url: "http://172.17.0.1:1337"
    api_token: "<api_token>"
env:
  HEXA_SERVER_URL: https://api.openhexa.org/
  HEXA_TOKEN: <token>
  HEXA_WORKSPACE: <workspace_slug>

[nazarfil]

I have added a commit to use newly added OpenHexaClient and changed test mocks accordingly.