BatchProcessor exception handling so that scheduler is not cancelled after connection error

[mmatloka]

See #143

[mmatloka]

@majst01 ??

[mmatloka]

@beckettsean ?

[beckettsean]

@mmatloka, let's wait a bit more for @majst01 to respond. InfluxData does not maintain this repo.

[majst01]

Hi,
still alive, sorry for the delay, im quite bussy.

Any chance to add a unit test to verify that your code works, other than that seems correct and obviously needed.

[andrewdodd]

This is essentially the same change that was supplied in #119 but with data loss instead of potential OOM.

I think there are some key 'feature' questions that need answers before you can really decide to include this. I have made a (unfortunately long) issue discussing this at #148.

[mmatloka]

@majst01 sure, I'll add the tests!

[mmatloka]

Hey @majst01,
I have added the test case.

[andrewdodd]

I've changed my mind. I think including this is reasonable (though I'll admit I'm wary of pokemon exception handling). I think it is a simple enough change to include for now.