Support NA, nan, and . as NaN values in various loaders

[r1cheu]

Summary

Support multiple representations of missing or invalid values (such as NA, nan, and .) as NaN values in various data loaders within the project.

Motivation

Currently, different datasets may use a variety of notations to indicate missing data. To enhance compatibility and reduce preprocessing friction, loaders should recognize NA, nan, and . as NaN values automatically.

Expected Behavior

• All relevant data loaders should interpret NA, nan, and . as NaN.
• Proper handling and conversion should be consistent throughout the library.
• Documentation should be updated to mention the supported missing value formats.