Replace t-test with chi-squared test.

The t-test was checking that the frequency mean matches the expected
value, but that test is instead looking at whether all values are
uniformly distributed.

It would be possible for values to have the right average but not a
uniform distribution. A chi-squared test is more suited for this.

tests/testthat/test-bitset.R

@@ -337,7 +337,7 @@ test_that("bitset is evenly sampled", {
       xs <- b$to_vector()
       freqs[xs] <- freqs[xs] + 1
     }
-    p <- t.test(freqs, mu=rate*N)$p.value
+    p <- chisq.test(freqs)$p.value
     expect_gt(p, threshold)
   }
 })