Refactor findSchemaElement to use named struct and add column_index

[unexge]

Summary

Problem: The findSchemaElement method returned a schema index that required users to subtract 1 to get the column index (as seen in the NYC taxi example: file.findSchemaElement(&.{name}).?.index - 1).

Solution:

• Define a named struct SchemaInfo instead of anonymous struct with fields:

  • column_index - index into the row group's columns array (leaf elements only)
  • max_definition_level - for handling nullability
  • max_repetition_level - for handling repeated elements
  • elem - the schema element itself

• Update findSchemaElement to track the column index by counting only leaf elements (elements without children) as it traverses the schema tree.

• Update the NYC taxi example and README to use column_index directly instead of index - 1.

• Add tests for:

  • Flat schema: verifies column indices 0, 1, 2 for columns foo, bar, ham
  • Non-existent paths: verifies null is returned
  • Optional columns: verifies correct definition/repetition levels
  • Nested schema: verifies correct column indices for nested struct fields (e.g., roll_num.min → 0, roll_num.variance → 5, PC_CUR.min → 6)