Enhance messy_linelist()

[joshwlambert]

This PR closes #191 and closes #192 by enhancing the messy_linelist() function.

The prop_missing setting in messy_linelist() now introduces missing values to <data.frame> cells not already NA, if the missing_value specified is NA (default). This replaces the random placement of missing values in the previous implementation of messy_linelist().

The .add_missing() function is added which contains a more sophisticated approach to inserting user-specified missing values (missing_value) into the line list <data.frame>.

• It samples elements of the <data.frame> to make missing depending on the missing_value and doesn't sample NA elements if the missing_value is NA.
• It performs type coercion in cases where the type of the missing_value and the type of the <data.frame> column differ to avoid unwanted coercions, e.g. numeric to Date.
• It records the type coercions and warns the user

int_as_word setting in messy_linelist() has been updated to prop_int_as_word to allow users to control the proportion of integer values that are convert to words using english::words().

inconsistent_id is added as a setting to messy_linelist(), which by default is off (FALSE), but when switched on appends random three letter prefixes or suffixes to a random ~10% sample of $ids.

Input checking for prop_missing and missing_value is added to messy_linelist().

New unit tests are added to test the behaviour of missing_value in messy_linelist() since adding the internal .add_missing() function.