add test cases for superfluous days

[TymekDev]

This PR adds cases for whole years and whole months to not have extra days, e.g. May (2023-05-11T00:00:00.00) to July (2023-07-11T00:00:00.00) being 2 months instead of current 2 months, 1 day.

Related issue: #262

[primer-css]

👋 Hello and thanks for pinging us! This issue or PR has been added to our inbox and a Design Infrastructure first responder will review it soon.

• 🎨 If this is a PR that includes a visual change, please make sure to add screenshots in the description or deploy this code to a lab machine with instructions for how to test.
• ⏩ If this is a PR that includes changes to an interaction, please include a video recording in the description.
• ⚠️If this is urgent, please visit us in #primer on Slack and tag the first responders listed in the channel topic.

[lukasoppermann]

I'd love for @keithamus to approve this as he is working on it. Fine from primer FR side

[keithamus]

Thanks, I'm looking into fixing this but I have higher priorities for now. A fix will come though!

[joshblack]

Sorry for the delay on my end, adding this to the first responder board to pick up 👀 @keithamus just wanted to check-in, what are the next steps for this?

[camertron]

I looked into this today and have become thoroughly confused. Are you specifying any precision argument anywhere?

If I try copy/pasting your test cases into example/index.html, it always gives me something like "1 year, 6 days, 4 minutes, 24 seconds." I don't understand why the tests fail the way they do when no precision is specified.

[TymekDev]

@camertron according to the README the precision defaults to second, so I'd assume that's how it works. I don't think it changes anything in this context, though.

I just put this into an HTML (using v4.4.4 via npm):

<relative-timeformat="duration" datetime="1944-12-13"></relative-time>

It showed "81 years, 2 months, 22 hours, 21 minutes, 45 seconds" for me. Even if I set the precision as high as year, then that would still be wrong—it should be 80 years. In other words, given a long enough duration, all precision levels are affected.

[francinelucca]

Thanks for opening this PR @TymekDev! sorry for lack of activity here. I've taken a look at the tests and looks like you're right, the code needs to be corrected. I don't want to get failing tests into main so I'm hoping we can land on a fix instead and then pull these in. I've dropped a comment in #262 (comment), hopefully we can work on this soon. Will leave this PR open until then.