feat: limit/offset on cog_gov_search() and cog_balances() (#57)

verbs were left materializing everything and slicing in R -- the pattern behind
the 2026-08-06 production incident. cog_gov_search() had no LIMIT at all, so an
unfiltered call returns the entire 40,336-row crosswalk.

Extracted the #39 machinery into R/pagination.R first (.validate_pagination(),
.paginate_sql(), .take_pagination_total()) rather than growing a third inline
copy: three definitions of what total_rows means is three places for it to
drift. Conflict refusals stay at the call sites because each verb's conflict
set differs. .verb_spendrev() now uses the shared helpers and is unchanged in
behaviour.

The empty-page fallback query is now passed as a thunk, so the unpaginated SQL
is only BUILT when an offset actually lands past the end instead of on every
paged call.

Two things #57 did not anticipate:

- cog_gov_search()'s ORDER BY was not a total order. population_acs DESC NULLS
  LAST leaves ties -- and the whole NULL block -- in scan order, so two requests
  can order them differently and a paged sweep duplicates one row while dropping
  another. Added canonical_govid as tiebreaker. Unpaginated output changes only
  in the relative order of already-tied rows.

- Basket mode returns one resolved row per requested name plus a sidecar
  covering all of them, so a page of it is not a page of anything the caller
  asked for. Refused with uscogdata_basket_pagination_conflict rather than
  silently ignoring the arguments.

Both default to NULL, so cog-api adopts them behind its existing formals()
probe with no lockstep deploy.

Suite: 1067 passed, 0 failed, 0 warnings (2 pre-existing live-corpus skips).

man/cog_balances.Rd

@@ -13,7 +13,9 @@ cog_balances(
   basis = c("harmonized", "raw"),
   recipe = NULL,
   state = NULL,
-  type = NULL
+  type = NULL,
+  limit = NULL,
+  offset = NULL
 )
 }
 \arguments{
@@ -76,6 +78,14 @@ wide era to the modern one.}
   and `govids_missing` are empty -- there is no id list to report against --
   and `provenance$scope$cohort` carries `state`, `type` and
   `n_governments` instead. A `govid`-named cohort reports exactly as before.}
+
+\item{limit}{Maximum number of result rows to return, pushed into the SQL
+rather than applied after materializing every row. `NULL` (default)
+returns everything. Cannot be combined with `recipe` -- see `offset` and
+`total_rows`.}
+
+\item{offset}{Rows to skip before `limit` starts counting (0-based).
+Ignored if `limit` is `NULL`; defaults to `0L` when `limit` is set.}
 }
 \value{
 Tibble with columns `year`, `canonical_govid`, `gov_name`,
@@ -94,6 +104,10 @@ Tibble with columns `year`, `canonical_govid`, `gov_name`,
   and `truncated` (the observed subtypes whose coverage falls short of the
   requested years). `expenditure_concept`/`revenue_concept` are `NA` --
   holdings are a stock, not a flow, so neither concept vocabulary applies.
+
+  When `limit` is set, also carries a `total_rows` attribute: the full
+  unpaginated row count, computed by the same query (`COUNT(*) OVER()`)
+  rather than a second scan.
 }
 \description{
 Returns Census cash-and-security holdings (`category_type = "balance"`):