Use internal checking function to reduce duplication

[Bisaloo]

• First commit creates an internal function containing checks common to all the cfr_() function
• Second commit uses a collection so that the input data is completely validated in one go, and an error message with all the issues is returned. This prevents the user from having to try and error multiple times, with a different error message each time.

[adamkucharski]

Thanks. Will review and merge into the latest 0.1.2 release PR #168 is complete (as seems like this is probably best sequence given CRAN submission was in limbo while this PR was added).

[adamkucharski]

I've reviewed these changes with a mis-specified data input column, i.e. ebola1976_first_30 <- rename(ebola1976_first_30,dates=date) and found the following issue:

The original check format stops at the assert_names step with an informative warning, preventing any further checks being evaluated:

checkmate::assert_names(
colnames(data),
must.include = c("date", "cases", "deaths")
)
Error: Assertion on 'colnames(data)' failed: Names must include the elements {'date','cases','deaths'}, but is missing elements {'date'}.

However, the new function seems to hit an error because it stores the output of this check along the way, but then reaches the assert_data_frame check, which can't be completed:

.check_input_data(data)
Error in `[.data.frame`(data, , c("date", "cases", "deaths")) : 
  undefined columns selected

[Bisaloo]

I see two potential fixes:

• keep the collection approach and make sure all errors can be collected without "catastrophic" failure. This would require complexifying the code a bit but I think it's doable
• revert to the previous approach of sequential checks & failures. This PR would thus only contain 734df58 (#161).

What would you prefer?