Add proper re-try behavior based 408, 400 and 503 errors and other responses

[GustaafL]

In the request docstring it mentions 408 explicitly. I think we should check for that here explicitly. Note that in the one test we have in which the server responds with a 408, the client actually gives up before the server sends something back.

Originally posted by @Flix6x in #7 (comment)

[ArdJonker]

In last days also http 502 and 504 (error) messages where received, also handle these?

[GustaafL]

I believe we should mention that this will re-try until a schedule is ready.

Actually, I'm not so sure this works as expected, as we seem to poll only on failure (408, 503), not on responses that tell us that the results are not ready - I believe those are coded as 400 right now.

@Flix6x should this be resolved in another issue, also in the flexmeasures server?
Or am I interpreting this wrongly?

Originally posted by @nhoening in #28 (comment)

[nhoening]

To be cleaer - It seems to me that 400 is the wrong code, so FlexMeasures server might require an issue, as well.

[Flix6x]

Note that the docstring here does not reflect the actual code here. The actual code does check whether the scheduling job is waiting or in progress, so for now we have that covered.

Related issue: FlexMeasures/flexmeasures#645