Format one-tuples and one-subscripts as multi-line if already multi-line

[bluetech]

This implement the proposal in #3918.

I wasn't sure if this needs to go as a preview option; it shouldn't affect already formatted code (so adheres to the Stability Policy), but it does change how unformatted input is formatted.

Split to multiple commits to make reviewing a bit easier. Can be squashed to a single commit. The actual meaningful change is the last commit.

I've used opening.lineno != closing.lineno as an easy way to detect "multiline". I think it's correct and simple but maybe it's not appropriate.

Tested on some of our projects and it has the intended effect. For skip-magic-trailing-comma = true configurations it should have no effect at all.

[github-actions]

diff-shades results comparing this PR (0a79e73) to main (58f31a7). The full diff is available in the logs under the "Generate HTML diff report" step.

╭─────────────────────── Summary ────────────────────────╮
│ 2 projects & 4 files changed / 18 changes [+13/-5]     │
│                                                        │
│ ... out of 2 514 425 lines, 11 770 files & 23 projects │
╰────────────────────────────────────────────────────────╯

Differences found.

What is this? | Workflow run | diff-shades documentation

[bluetech]

I started looking into the SQLAlchemy diff-shades stability failure. I minimized it to the following failure:

Mode(target_versions={<TargetVersion.PY38: 8>}, line_length=79, string_normalization=True, is_pyi=False, is_ipynb=False, skip_source_first_line=False, magic_trailing_comma=True, experimental_string_processing=False, python_cell_magics=set(), preview=True)
--- source
+++ first pass
@@ -1,8 +1,6 @@
 if True:
     if True:
         if ooooooooooooooooooooooooo and oooooooooooooooooo[ooo].oooooooo == (
             0,
         ):
-            Baz(
-                ""
-            )
+            Baz("")
--- first pass
+++ second pass
@@ -1,6 +1,6 @@
 if True:
     if True:
-        if ooooooooooooooooooooooooo and oooooooooooooooooo[ooo].oooooooo == (
-            0,
-        ):
+        if ooooooooooooooooooooooooo and oooooooooooooooooo[
+            ooo
+        ].oooooooo == (0,):
             Baz("")

However when I try to understand what happens in the second pass by running black on the output of the first pass, I get no changes. So I'm a bit stumped; maybe I misunderstood how it works and the second pass is not just running black again on the output of the first pass? I'll try figure it out, in the meantime I'll mark the PR as draft.

[bluetech]

Will close this for now, might try again in the future.