Prevent stack deployment from timing out

[vikte]

When deploying stacks on machines that are either slow/low resource or have slow internet connection, the deployment step often takes longer than 10 seconds. When that happens the deploy progress popover will show an error TypeError: Error in input stream.

This happens because the connection to deploy-stream endpoint gets eventually timed out either by any proxies in between or bun as it uses the default IDLE_TIMEOUT env var value of 10 seconds. Then the response reader errors out and produces the error TypeError: Error in input stream.

This PR adds a heartbeat to the deploy-stream API that fires every 5 seconds. The frontend will filter it out as the actual message doesn't start with data: . As such this is purely to keep the connection open and doesn't change or add anything visible to the user. The implementation is very similar to how e.g. merged-logs API already does it.

Additionally the hawser edge requests' timeout has been extended from 30 seconds to 2 minutes (when non-streaming). 30 seconds in mentioned scenarios is not enough and the popover would show error Connection timeout. 2 minutes is something that worked in my scenario, but happy to change to a more optimal value if needed.

I think this potentially addresses #329