Add KDocs for non-deprecated `percentile` APIs

Add KDocs for non-deprecated percentile APIs

The file core/src/main/kotlin/org/jetbrains/kotlinx/dataframe/api/percentile.kt holds the public percentile APIs, and the coverage sheet marks the group percentile/* as missing KDocs. None of these functions are currently documented (there is only a TODO KDocs comment block at the top of the file). Please add concise KDocs for the current non-deprecated APIs only, excluding deprecated and binary-compatibility (@AccessApiOverload) overloads.

Scope

APIs needing KDocs, by receiver (each takes the percentile: Double argument):

• DataColumn.* — percentile, percentileOrNull, percentileBy, percentileByOrNull, percentileOf, percentileOfOrNull
• DataRow.* — rowPercentileOf, rowPercentileOfOrNull
• DataFrame.* — percentile, percentileOrNull, percentileFor, percentileOf, percentileOfOrNull, percentileBy, percentileByOrNull
• Grouped.* — percentile, percentileFor, percentileOf
• GroupBy.* — percentileBy
• Pivot.* — percentile, percentileFor, percentileOf, percentileBy
• PivotGroupBy.* — percentile, percentileFor, percentileOf, percentileBy

KDocs should explain

• What each operation returns:
  • percentile — the percentile value (over the whole column / frame);
  • percentileFor — a DataRow/DataFrame with a separate percentile per selected column;
  • percentileOf — the percentile of a value computed per row/element via the given expression;
  • percentileBy — the whole row (or reduced group / pivot) at the given percentile of a comparable selector;
  • the *OrNull variants — the same, but return null instead of failing on empty / all-null input;
  • the grouped / pivoted variants — computed per group / per pivot cell.
• The type behavior (already captured in the TODO KDocs block, lines 31–46):
  • for primitive numbers the result is Double (or null);
  • for other Comparable types the result is a value of that same type (or null);
  • careful: non-primitive numbers follow the comparable rules, not the numeric ones;
  • the non-OrNull variants fail via suggestIfNull on empty / all-null input; the *OrNull variants return null.
• The percentile parameter (a Double, e.g. 50.0 for the median) and how skipNaN treats NaN in floating-point columns.
• The quantile estimation method used: QuantileEstimationMethod.R8 for primitive numbers, R3 otherwise; percentileBy uses R3.
• The KT-76683 caveat: for the lambda-return-type percentileOf / percentile { } overloads on non-number comparable types, explicit type arguments are required — e.g. df.percentile { intCol } works, but a String column needs df.percentile<_, String> { stringCol } (or passing the lambda positionally). This must be shown in the KDoc.
• The optional name parameter for the Grouped variants and the separate parameter for the Pivot / PivotGroupBy variants.

Acceptance criteria

• Concise KDocs are added to all non-deprecated percentile / percentileOrNull / percentileFor / percentileOf / percentileOfOrNull / percentileBy / percentileByOrNull / rowPercentileOf / rowPercentileOfOrNull overloads in percentile.kt.
• No documentation is added for deprecated / @AccessApiOverload overloads (including the DeprecationLevel.ERROR rowPercentile() / rowPercentileOrNull() stubs).
• KDocs cover the percentile and skipNaN parameters, the primitive-number → Double vs comparable → same-type return rules, and the empty/all-null behavior (suggestIfNull vs null).
• KDocs document the KT-76683 explicit-type-argument requirement for the non-number comparable lambda overloads.
• KDocs mention the quantile estimation method (R8 for numbers, R3 otherwise / for percentileBy).
• KDocs clearly distinguish the value-returning, per-column aggregation (percentileFor), and row-selecting (percentileBy) variants.
• Docs are consistent with the site docs at percentile.html.
• The TODO KDocs comment block at the top of the file is removed once its content has been folded into the actual KDocs.