Allow for a tolerance in assert_daily_data() to enable NAs in count data

[kaitejohnson]

Goal

Currently the stan model/ preprocessing can handle this (as we did this in practice last year to exclude specific days of data), however we set up a very strict structure for assert_daily_data() which checks that all days are present in the date vector of the admits data. We should relax this a bit so that there can be some amount of missingness in the admissions.