Improve long values in dict literals

[yilei]

Description

This PR changes the preview style:

• Wraps long values in dict literals in parentheses. Fixes Black should add parentheses to long values in dict literals #620. Fixes Line too long: Lambda in dictionary value #808.
• Correspondingly, parentheses around short values in dict literals are removed.
• In preview string processing, long lambda string values are now wrapped in parentheses.
• Wrapping dict values happens to fix Black --preview mode produces invalid code in the first pass, causing second pass to fail #3117 where implicitly concatenated r-string dict values crash in string processing. Fixes Black --preview mode produces invalid code in the first pass, causing second pass to fail #3117

Usually I'd like to keep the PR's scope minimal, but the lambda change & dict value change are depended by each other so it's hard to split into two PRs.

Checklist - did you ...

• Add an entry in CHANGES.md if necessary?
• Add / update tests if necessary?
• Add new / update outdated documentation?

[github-actions]

diff-shades results comparing this PR (ba1b968) to main (a282181). The full diff is available in the logs under the "Generate HTML diff report" step.

╭───────────────────────── Summary ─────────────────────────╮
│ 10 projects & 99 files changed / 1089 changes [+608/-481] │
│                                                           │
│ ... out of 2 355 274 lines, 10 989 files & 23 projects    │
╰───────────────────────────────────────────────────────────╯

Differences found.

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

[JelleZijlstra]

Notes from diff-shades:

             queryset = queryset.extra(
                 select={
                     "_prefetch_related_val_%s"
-                    % f.attname: "%s.%s"
-                    % (qn(join_table), qn(f.column))
+                    % f.attname: "%s.%s" % (qn(join_table), qn(f.column))
                     for f in fk.local_related_fields
                 }
             )

Note that both the key and the value in the dictcomp use % formatting. Looks a little weird, maybe we should put the key and value on separate lines. Seems out of scope for this PR though.

Pandas has a similarly odd-looking case with a + in the key:

         tests = [
             (
                 BusinessHour(),
                 {
                     Timestamp("2014-07-04 15:00")
-                    + Nano(5): Timestamp("2014-07-04 16:00")
-                    + Nano(5),
+                    + Nano(5): Timestamp("2014-07-04 16:00") + Nano(5),
                     Timestamp("2014-07-04 16:00")
-                    + Nano(5): Timestamp("2014-07-07 09:00")
-                    + Nano(5),
+                    + Nano(5): Timestamp("2014-07-07 09:00") + Nano(5),
                     Timestamp("2014-07-04 16:00")
-                    - Nano(5): Timestamp("2014-07-04 17:00")
-                    - Nano(5),
+                    - Nano(5): Timestamp("2014-07-04 17:00") - Nano(5),
                 },
             ),

Otherwise I like the changes from the PR: they make it easier to see when one dict key ends and the next one begins, because values get indented.

[JelleZijlstra]

thanks, just noticed two typos

[yilei]

Thanks for the review! Agreed above two examples are out of scope from this PR, they concern the dict keys. Opened #3442 separately.