feat: cog_spending + cog_revenue + cog_explain

Three core query verbs over the spending_annotated / revenue_annotated
DuckDB views. Each verb accepts vector govid, vector years, optional
category filter, per_capita flag, and adjust_to_year for CPI-U
real-dollar conversion (bundled index).

Amounts are returned in full USD (SUM(amt) * 1000) so callers can
freely rescale to millions/billions. The $1,000s -> $USD conversion
is recorded in provenance$transformations$units_conversion.

Every result carries an attr(., "provenance") list matching
inst/schemas/provenance-v1.json. cog_explain() prints the structured
form via cli or returns the raw list for MCP/JSON consumers.

Also: .fetch_or_cache_manifest() now handles local fixture paths so
tests can point USCOGDATA_FIXTURE_URL at the pipeline publish_cache/
without a working HTTP server.

Tests: 80 pass / 0 fail. devtools::check() 0E/0W/2N (both notes
pre-existing / environmental).

tests/testthat/test-explain.R

@@ -0,0 +1,32 @@
+test_that("cog_explain prints verb header and target", {
+  skip_if_no_corpus()
+  r <- cog_spending("101006006", 2020L, "Corrections")
+  # cli writes to stderr; capture both stdout and message streams.
+  txt <- paste(c(
+    capture.output(cog_explain(r)),
+    capture.output(cog_explain(r), type = "message")
+  ), collapse = "\n")
+  expect_true(grepl("cog_spending", txt))
+  expect_true(grepl("Corrections", txt))
+  expect_true(grepl("101006006", txt))
+})
+
+test_that("cog_explain format='list' returns structured provenance", {
+  skip_if_no_corpus()
+  r <- cog_spending("101006006", 2020L, "Corrections")
+  prov <- cog_explain(r, format = "list")
+  expect_identical(prov, attr(r, "provenance"))
+})
+
+test_that("cog_explain returns result invisibly for chaining", {
+  skip_if_no_corpus()
+  r <- cog_spending("101006006", 2020L, "Corrections")
+  res <- withVisible(cog_explain(r))
+  expect_false(res$visible)
+  expect_identical(res$value, r)
+})
+
+test_that("cog_explain errors on non-verb input", {
+  df <- tibble::tibble(a = 1)
+  expect_error(cog_explain(df), "provenance")
+})