133 fix residuals

[dajmcdon]

Closes #133.

The basic issue is that the test data must be complete (or at least have enough non-NAs to do something). In the sliding vignette, many locations are missing recent percent_cli so the predictions got a bit odd.

Changes:

• Update to get_test_data() to allow for NA fill and print messages if there's insufficient data for the fill.
• Move epiprocess to Depends, since this package is fundamentally linked to it.
• Changed default residual_quantile values (the old ones were strange).

[rachlobay]

After fixing that one test & adding a little more description to n_recent in get_test_data, looks good to merge (after all checks pass).