Add coveralls to travis

[JoranAngevaare]

What is the problem / what does the code in this PR do
Add https://docs.coveralls.io/ to travis testing.

Can you briefly describe how it works?
Just like in strax

My concern
I'm just copy pasting code here. I have added coveralls for straxen: https://coveralls.io/github/XENONnT/straxen, however I'm not sure if this is going to work the first time. If someone is willing to give me the green light I can just push to master.

I think this would be great to add because it shows how good/bad your PR is and increases the incentive to write tests.

[JoranAngevaare]

@WenzDaniel , are we running into AxFoundation/strax#328 at line https://github.com/AxFoundation/strax/blob/master/strax/processing/pulse_processing.py#L116 at commit 77b2e18?

This just happens for numbaless coverage (newly added in this PR) so you probably have never seen it before.

[WenzDaniel]

This is interesting, but no the Issue you are linking would be the opposite I think. Back then I had the issue that my assignment

data = thing['data'][thing['length']]

was a view onto thing['data']. Hence when I modified data, I indirectly modified thing['data'] which was not desired. Hence this read-only thing you are describing here would have saved me from this issue.
I know that one can make an array immutable via array.flag.writeable = False, but I do not think that we are explicitly doing this somewhere. Further, I checked the requirements since v0.12.0 and numpy and numba has not change since then. Have you updated your env recently?

[WenzDaniel]

The thing I do not understand though, why is it just popping up for the nveto_recorder?

Okay, I understand why, funny. Here is what happened and now happens:

The arrays we are returning in the compute method of our plugins are read-only which makes sense. In the past there was a bug in numba that a njited function which work on a immutable structure array would just do nothing (see numba/numba#5950). However, it looks like that they have changed this recently and now numba throws an error.

Since the function was not doing anything before, and I do not see that it would make any difference, we can just remove it.

[JoranAngevaare]

Thanks @WenzDaniel , trying to work from here now

[JoranAngevaare]

Ok, seems to work :) @ershockley or @WenzDaniel willing to review (is just technical so might do without). Maybe to save some time on travis tests, I will disable py3.7 testing and just to numbaless py3.7. Any opinions?

@daniel, 29% oopf, we have work to do..

[JoranAngevaare]

Ok, changed the way it looks at the package, it's now at 61 percent (apparently this test_plugins is very effective). I will continue and merge.