Extend the mostly logic to define success using absolute row numbers in addition to percentage

[wllgrnt]

Is your feature request related to a problem? Please describe.
In our tables we have a fixed number of rows that fail the column expectations for historic reasons, but the tables grow all the time so the required mostly percentage to ensure new bad data is caught shrinks every day. If we could extend the idea behind themostly kwarg to allow me to set a number of rows I expect to fail (i.e. success is when unexpected_count > bla, I think) then we could do much more stringent testing.

Describe the solution you'd like
Two obvious options would be:

1. Check if mostly is a float [0,1). If so, run the normal mostly logic. Otherwise if mostly is an int, interpret it as the maximum number of unexpected rows to allow. (Otherwise handle the invalid argument as per usual).

2. Add a new kwarg called e.g. mostly_absolute and check that the function caller doesn't try to use both.

I defer to the core devs on which they'd prefer, happy to submit a PR once there's consensus on the best signature!

[eugmandel]

@wllgrnt Thank you for posting this idea! I want to make sure that more team members get to see it before we decide to move on it.

[eugmandel]

@jcampbell I think it is a good enhancement - thoughts?

[vikramaditya91]

+1 on this request