GNU testsuite comparison:

Skip an intermittent issue tests/shuf/shuf-reservoir (fails in this run but passes in the 'main' branch)
Skip an intermittent issue tests/sort/sort-stale-thread-mem (fails in this run but passes in the 'main' branch)
Skip an intermittent issue tests/tail/inotify-dir-recreate (fails in this run but passes in the 'main' branch)

GNU testsuite comparison:

Congrats! The gnu test tests/seq/seq-epipe is now passing!

GNU testsuite comparison:

Note: The gnu test tests/env/env-signal-handler is now being skipped but was previously passing.

I've spent a bunch of time prototyping approaches to the issue of locale and calendar support in date and I am aligned that the approach is to pre-intercept and do our own replacement before passing it to Jiff. Its unfortunate that it means that we essentially have to recreate the majority of the date parser. That being said reviewing this is a bit tricky because its going to be a long journey to get to full compatibility with this approach and we could go back and forth for a long time with things that don't match the gnu date implementation so I'm going to try to keep it high level. From a review perspective I think it would be better to build forward and merge as is and keep iterating on it to keep the PR's smaller and keep adding more regression test cases.

The high level comments I would have mainly relate to the approach of the placeholder, when testing locally I could not come up with an example where doing the injection before passing it to Jiff would create a conflict compared to injecting the placeholders then doing the replacements after Jiff. I think it would be simpler to just perform the replacement in advance and not have to worry about adding the placeholder, but if you came across a scenario where that caused an issue this approach would make sense.

The other high level comment is related to using jiff-icu. In my prototyping I mainly used jiff-icu to get the locale data:

  let formatted = formatter.format(&date).to_string();
  // Extract month name from formatted date like "15 janvier 2000" or "2000-01-15"
  // Look for a word that contains letters (the month name)
  let words: Vec<&str> = formatted.split_whitespace().collect();
  words.iter()
      .find(|word| word.chars().any(|c| c.is_alphabetic()))
      .map_or_else(String::new, |s| (*s).to_string())

The Jiff ICU library makes the code to do this less flaky and more explicit:

  let icu_zdt = ZonedDateTime::convert_from(&zdt);

  let formatter = DateTimeFormatter::try_new(
      locale!("fr-FR").into(),
      fieldsets::YMDET::long(), 
  )?;
  formatter.format(&icu_zdt).to_string()

Would you be okay if I merge this as is, from the context that there is so far to go and the first PR will be the largest and just adding the locale env variable parser and the dependencies we need already puts us in the right direction and we can make the implementation way more robust in the future.

Would you be okay if I merge this as is

sounds good!