Deal with dist_quantiles() CHECK issue

[brookslogan]

• address the deprecation, migrating to hardhat fn
• address the error if it persists after migration

── Warning ('test-compactify.R:124:3'): compactify works on distributions ──────
`dist_quantiles()` was deprecated in epipredict 0.1.11.
i Please use `hardhat::quantile_pred()` instead.
Backtrace:
    ▆
 1. ├─tibble::tibble(...) at test-compactify.R:124:3
 2. │ └─tibble:::tibble_quos(xs, .rows, .name_repair)
 3. │   └─rlang::eval_tidy(xs[[j]], mask)
 4. └─epipredict::dist_quantiles(c(3, 5, 9), c(0.2, 0.5, 0.8))

══ Failed tests ════════════════════════════════════════════════════════════════
── Error ('test-compactify.R:124:3'): compactify works on distributions ────────
<vctrs_error_ptype2/vctrs_error_incompatible_type/vctrs_error_incompatible/vctrs_error/rlang_error/error/condition>
Error in `vec_c(...)`: Can't combine `..1` <quantiles> and `..4` <distribution>.
Backtrace:
     ▆
  1. ├─tibble::tibble(...) at test-compactify.R:124:3
  2. │ └─tibble:::tibble_quos(xs, .rows, .name_repair)
  3. │   └─rlang::eval_tidy(xs[[j]], mask)
  4. ├─base::c(...)
  5. ├─vctrs:::c.vctrs_vctr(...)
  6. │ └─vctrs::vec_c(...)
  7. └─vctrs (local) `<fn>`()
  8.   └─vctrs::vec_default_ptype2(...)
  9.     ├─base::withRestarts(...)
 10.     │ └─base (local) withOneRestart(expr, restarts[[1L]])
 11.     │   └─base (local) doWithOneRestart(return(expr), restart)
 12.     └─vctrs::stop_incompatible_type(...)
 13.       └─vctrs:::stop_incompatible(...)
 14.         └─vctrs:::stop_vctrs(...)
 15.           └─rlang::abort(message, class = c(class, "vctrs_error"), ..., call = call)

[ FAIL 1 | WARN 7 | SKIP 0 | PASS 1340 ]
Error: Error: Test failures
Execution halted

2 errors ✖ | 1 warning ✖ | 1 note ✖
Error: Process completed with exit code 1.

[brookslogan]

Unfortunately, hardhat::quantile_pred() doesn't seem to cleanly support vec_c with NA in downstream ops, unlike the old backend of dist_quantiles() (though in tests, I have some explicit distributional::dist_missing call... did vec_c also have issues with casting NA there? I don't know if there's any equivalent to that in hardhat though). I can't immediately tell if this is a hardhat issue or vctrs issue, and am trying to work through trouble installing winch for more details / bug filing.

[brookslogan]

• The test needs updated to directly use pred_quantiles(), not use dist_missing(), and to not have level-changing.
• Add specialized epix_slide for epi_slide_opt #611 appears to incidentally deal with as_epi_archive() running into issues due to dplyr::lag() creating a not-well-handled NA.
• Need to report these NA issues upstream.