Add KDocs for public reverse APIs

[zaleslaw]

Add KDocs for public reverse APIs

Closes #1980. All five public reverse overloads now have KDocs. No behavior changes.

KDoc

• Added a shared KoDEx template CommonReverseDocs in api/reverse.kt. It holds the first
  line, the website link, @param [T] and @return. Each overload only sets its own parts
  (RECEIVER, UNIT, DETAILS, SEE_ALSO, TYPE_PARAM). Same pattern as CommonTakeAndDropDocs.
• Added DocumentationUrls.Reverse for the reverse.html link.
• The text describes the contract only — what you get back, not how it is done.

What each overload now states:

Overload	Documented contract
DataFrame	Only row order changes. Columns, types and schema stay the same.
DataColumn	Name, type and kind stay the same. The result is typed as DataColumn, so for a column group you need asColumnGroup().
ColumnGroup	Name and nested structure stay. Values in all nested columns stay aligned row-wise.
FrameColumn	Only the order of the dataframes changes. Rows inside each dataframe keep their order.
ValueColumn	Name and type stay the same.

Tests

3 tests -> 10. Every statement in the KDocs now has a test. The three old tests are unchanged.

The old tests did not call the overloads one may expect. columnOf(a, b) returns
DataColumn<DataRow<*>>, not ColumnGroup. columnOf(1, 2, 3) returns DataColumn<Int>, not
ValueColumn. So ColumnGroup.reverse() and ValueColumn.reverse() were never called by any
test. The new tests build these columns explicitly with DataColumn.createColumnGroup() and
DataColumn.createValueColumn().

New tests:

• dataframe keeps schema and size — schema and row count.
• column keeps name and type.
• column overload keeps the column kind — a group stays Group, a frame stays Frame, a value
  stays Value. The DataColumn<*> types in this test are required: without them the calls go to
  the other overloads and the test would check something else.
• column group type can be restored after the column overload — asColumnGroup() round-trip.
  ColumnGroup is not a subtype of DataColumn (it extends BaseColumn and DataFrame), so this
  cast is the only way to get the group type back.
• column group reverses rows as a whole — real ColumnGroup receiver. It checks the name, the
  schema and the row pairs. It compares against explicit expected pairs and not against
  DataFrame.reverse(): comparing two implementations can pass while both are wrong.
• frameColumn — the frames swap places, rows inside them keep their order.
• valueColumn — the ValueColumn<Int> type of the result is the assertion here. The compiler
  checks it.

Note on assertions: column equality (impl/columns/Utils.kt) compares name, type and values. One
shouldBe on a whole column therefore already covers name and type. Separate name() / type()
asserts are used only where there is no whole-column comparison.

KDoc guidelines

Two notes added to KDOC_GUIDELINES.md, both found while writing the template:

• A multi-line @set value keeps the indent of its continuation lines. An indent of 4+ spaces
  becomes a code block in Markdown, so the paragraph renders as monospace.
• @param and @return must stay at the end of a KDoc. Text cannot be added after @include, so
  all parts that differ between overloads must be passed in as @set arguments.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[Jolanrensen]

So, like:

/**
 *  {@get A}
 * {@set A   hello}
 */

?
I don't think this can happen, because the argument in @set is trimmed

[zaleslaw]

Ah, okey, agent faced several time with the problem and decided to include it, probably need to double-check

[Jolanrensen]

can you give an example where this holds?

[Jolanrensen]

When using a tag on a newline, no {} are needed

[Jolanrensen]

nowhere in this kdoc are {} necessary

[zaleslaw]

Will it work? Was it in the initial KDOC or KODEX guide?

[Jolanrensen]

Yes, it works; block tags end when another block tag is encountered. I'm not sure how explicitly it's described, but it's how all other block tags work :)

[Jolanrensen]

Brackets are only needed when you want to put a tag inline

[Jolanrensen]

Normally you can see where a tag starts and ends using the intellij plugin of KoDEx, but it seems that has broken in the latest updates...

[Jolanrensen]

oxford comma missing

[Jolanrensen]

The produced KDocs look good and useful :)