Fix start_date and end_date never being set in calculated totals

[arpitjain099]

There are two bugs in the start/end date block in src/process_results/result_totals_calculator.js, and they hide each other, so they have to be fixed together.

The guard is if (result.data[0].data). Result rows don't have a data key, they have date plus the metric keys, so the guard is always false and start_date / end_date are never set. Right inside that branch, result.date[0].date should be result.data[0].date. result.date is undefined, so if the guard ever passed, this would throw.

That's why fixing only the guard is worse than leaving it alone: it turns a silent no-op into a TypeError. I checked this rather than assuming it. Correcting the guard on its own takes the suite from 1 pre-existing failure to 19, all TypeError: Cannot read properties of undefined (reading '0').

On intent, I didn't want to guess, so I traced it. Both typos came in together in 55a6f79, when this moved out of process-ga-data.js. The code before that move was:

if ((result.data.length > 0) && (result.data[0].date)) {
    if (result.data[0].date == "(other)")
        result.totals.start_date = result.data[1].date;
    ...

So result.data[0].date in both spots is the original behavior, and the result.data.length > 0 part is already covered by the early return at the top of calculateTotals. This restores that and nothing more.

One knock-on: test/actions/query_google_analytics.test.js asserted a totals object with no dates in it. That fixture is 24 hourly rows all on 20170130, so it now gets start_date and end_date of 2017-01-30. I updated the expectation. Happy to hear if any consumer would rather not see these fields appear, since they've effectively been absent for years.

Tests added in test/process_results/result_totals_calculator.test.js: dates set from first and last row, a (other) first row skipped in favor of row 1, and rows with no date dimension still not getting the fields.

Verified on Node 22 (per .nvmrc) with NODE_ENV=test npx mocha:

• before: 212 passing, 1 failing
• both bugs reverted, new tests in place: 213 passing, 3 failing (the 2 new date tests plus the pre-existing one)
• guard fixed, typo left in: 174 passing, 19 failing, all TypeError
• with this change: 215 passing, 1 failing

The one constant failure is PostgresPublisher, which is ECONNREFUSED against a database I don't have locally. It fails identically on an unmodified checkout. npx eslint . and prettier --check are both clean.