Add KDocs for `DataRowApi` functions

[zaleslaw]

Add KDocs for DataRowApi functions

The file core/src/main/kotlin/org/jetbrains/kotlinx/dataframe/api/DataRowApi.kt is partially documented. Add concise KDocs for the non-deprecated members that still lack them, excluding deprecated / @AccessApiOverload overloads.

Already documented (leave as-is)

• diff (all 4 overloads) and diffOrNull (all 4) — shared anchors DiffDocs / DiffOrNullDocs.
• prev(), next().

⚠️ Bug to fix while here: the next() KDoc @return line is wrong — it says "Returns the previous [DataRow] … or null if the current row is the last" (copy-pasted from prev()). Correct it to "next".

Scope — not documented (grouped)

• Row values — valuesOf<R>(), namedValues(), namedValuesOf<R>(), getValue<T>(columnName), getValueOrNull<T>(columnName).
• NameValuePair schema — the NameValuePair data class and the name / value accessor properties on ColumnsContainer<NameValuePair<*>> / DataRow<NameValuePair<*>>.
• Membership — containsKey(columnName), contains(column) operator.
• Row metadata — columnsCount(), columnNames(), columnTypes().
• Navigation / windowing — relative(Iterable<Int>), relative(IntRange), movingAverage(k, expression).

Out of scope: deprecated isEmpty / isNotEmpty, the deprecated getRow / getRows / getRowOrNull shortcuts, and all deprecated ColumnReference / KProperty overloads of getValue / getValueOrNull / containsKey / contains.

KDocs should explain

• getValue vs getValueOrNull: both look up a value by column name and cast to T; getValue throws if the column is missing, getValueOrNull returns null. Note the unchecked cast to T.
• valuesOf vs namedValues / namedValuesOf: valuesOf<R>() returns the row's values filtered to type R; namedValues() returns all (name, value) pairs as NameValuePair<Any?>; namedValuesOf<R>() returns only the pairs whose value is R.
• NameValuePair: it is the result schema of transpose / namedValues*; document the name / value accessors and why they exist (so row.transpose().name / .value resolve).
• Metadata accessors: columnsCount / columnNames / columnTypes describe the owning DataFrame's columns as seen from the row.
• relative: returns a sub-DataFrame of rows at offsets relative to the current row index, clamped to valid bounds (out-of-range offsets are dropped). movingAverage(k, ...) averages the expression over the current and previous k-1 rows.

Recommendations — make these KDocs genuinely useful (for users and for agents reading the source)

1. First sentence states the contract, standalone — what it returns and the key condition; it's what IDE quick-doc and an agent's context both read first.
2. Always give @throws / explicit-null behavior — pair getValue (throws) with getValueOrNull (returns null) so the reader picks the right one without reading the body.
3. Disambiguate the value-listing trio (valuesOf / namedValues / namedValuesOf) by what is filtered and what is returned; one line each beats generic prose.
4. Cross-link with @see the pairs and relatives: getValue↔getValueOrNull, prev↔next, relative↔movingAverage, namedValues↔transpose/NameValuePair.
5. One minimal example per family using a canonical df row (e.g. df[0].getValue<Int>("age")), showing the resulting type in a comment.
6. Name concrete types with links ([NameValuePair], [DataRow], [KType]) rather than "a value"/"a pair".
7. Call out non-obvious edges the signature hides: the unchecked as T cast in getValue*; relative clamping/dropping out-of-range offsets; movingAverage dividing by the actual count near the start of the frame.
8. Don't restate the signature — add the semantics types can't express.

Acceptance criteria

• All members listed under Scope have concise KDocs; the already-documented diff / diffOrNull / prev / next are preserved.
• The next() @return copy-paste bug is fixed.
• getValue / getValueOrNull document throw-vs-null and the unchecked cast; the value-listing trio is clearly disambiguated.
• Related members are @see-linked.
• No documentation is added for deprecated / @AccessApiOverload overloads.