ustfd 0.4.3

Better documentation of how filters work on API requests.
Improve documentation and use roxygen reuse functionality to reduce dupes
Refactor parsed_payload to eliminate dplyr dependencies
Refactor ustfd_all_pages in an attempt to reduce memory use. Previously, request aggregation was temporarily using more memory than necessary and requests resulting in a very high number of records (>200,000) were leading to large temporary memory allocations, potentially leading to performance issues or, in rare cases, crashes.

ustfd 0.4.2

During the week ending August 18th the API stopped coding record numbers as data-type NUMBER and started coding them as INTEGER in the meta record. The INTEGER type did not previously exist and therefore did not have an associated mapping.

removed ustfd_endpoints() and ustfd_field_dictionary(). They are replaced by ustfd_datasets(), ustfd_tables(), and ustfd_table_columns()

the dictionaries can now be easily updated before release instead of relying on a static endpoint list.
functions that had NA defaults now have NULL defaults.

Better error handling using rlang
YEAR, QUARTER, MONTH, and DAY columns are treated as integers, not numeric
PERCENTAGE columns no are no longer multiplied by 0.01. This change breaks existing behavior but was absolutely necessary to do before anyone relied on it.
Previously, a request that generated zero results would result in an error
ustfd_response_payload now guarantees the return is a tibble with the correct number of columns in the correct types, even if a the request generates no results.