Fix a mistake in the docstring of sem (#917)

* Fix a mistake in the docstring of `sem`

* Pass the Codecov token as an input (instead of an environment variable)

* Codecov: set `fail_ci_if_error` to `false` iff the PR is from a fork

* Whitespace

.github/workflows/ci.yml

@@ -38,9 +38,21 @@ jobs:
       - uses: codecov/codecov-action@v4
         with:
           file: lcov.info
-          fail_ci_if_error: true
-        env:
-          CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
+          # This is a public repo.
+          # We are using v4 of the codecov-action.
+          # If the PR is from a fork, then Codecov allows us to use
+          # tokenless Codecov uploads.
+          # If the PR is not from a fork, then Codecov does not allow
+          # us to use tokenless Codecov uploads, and thus we must use
+          # the token.
+          # Tokenless Codcov uploads are prone to random failures,
+          # due to Codecov's servers hitting GitHub rate limits.
+          # Therefore:
+          # 1. If this CI run is not a PR run, we set `fail_ci_if_error` to `true`.
+          # 2. If this CI run is a PR run and the PR is not from a fork, we set `fail_ci_if_error` to `true`.
+          # 3. If this CI run is a PR run and the PR is from a fork, we set `fail_ci_if_error` to `false`.
+          fail_ci_if_error: ${{ (github.event_name != 'pull_request') || (github.repository == github.event.pull_request.head.repo.full_name) }}
+          token: ${{ secrets.CODECOV_TOKEN }}
   docs:
     permissions:
       contents: write

src/scalarstats.jl

@@ -431,7 +431,7 @@ Return the standard error of the mean for a collection `x`.
 A pre-computed `mean` may be provided.
 
 When not using weights, this is the (sample) standard deviation
-divided by the sample size. If weights are used, the
+divided by the square root of the sample size. If weights are used, the
 variance of the sample mean is calculated as follows:
 
 * `AnalyticWeights`: Not implemented.