Stricter input checking for proportion_transmission()

[joshwlambert]

This PR tightens up the input checking in proportion_transmission().

proportion_transmission() now errors when:

• the percent_transmission argument is equal to 0 or 1. Previously percent_transmission could be greater or equal to 0 and lesser than or equal to 1. When equal to 0 or 1, the function would produce invalid results (e.g. NaN) or would not produce sensible results due to numeric imprecision (e.g. percent_transmission = 0 when method = "t_20" was > 0). When percent_transmission was equal 1 the function would error when method = "t_20" and results for method = "p_80" would vastly differ depending on if simulate was TRUE or FALSE.
• k is equal to 0. This produce errors weird behaviour similar to those outlined for percent_transmission.

Infinite values of k are approximated by setting them to 1e5. This was again due to errors and invalid results due to the inability of the internal proportion_transmission() methods not being able to work with Inf values.

The formatting of the proportion_transmission() and proportion_cluster_size() (when format_prop = TRUE which is the default) is now printed to 3 significant figures, rather than rounding, as values close to zero were being rounded to zero which could mislead users.

A new %gt% operator is added to the package to provide a succinct way to check if numeric variables are above a certain value and if so set than to the value with a message to the user. This operator is used in .prop_transmission_t20().

Unit tests are added to check the new input checking and variable setting, as well as for %gt%.