updated documentation regarding behavior of rbindlist when applied to difftime objects with different units

[Nj221102]

closes #4541

Summary

The current implementation of rbindlist does not preserve the units of difftime objects correctly when combining data.table or data.frame objects with different units. Instead, it converts all difftime values to a common unit without adjusting the values themselves. This can lead to incorrect time differences in the resulting data.table.

This pull request proposes updating the documentation to inform users that they need to manually ensure consistent units for difftime objects before using rbindlist to avoid inaccuracies.

[ben-schwen]

As pointed out by Jan in the issue I don't think that a tailored solution to a certain class is a favorable solution. Adding documentation about a single class is therefore not ideal either.

Can we generalize this to classes where the unit depends on class attributes?

[MichaelChirico]

I agree about the need to be more generic here.

[Nj221102]

@ben-schwen is this good enough ??

When binding lists of \code{data.table} or \code{data.frame} objects containing objects with units defined by class attributes (e.g., \code{difftime} objects with different units), the resulting \code{data.table} may not preserve the original units correctly. Instead, values will be converted to a common unit (typically the smallest unit present in the input) without proper conversion of the values themselves. This issue applies to any class where the unit or precision is determined by attributes. Users should manually ensure that objects with unit-dependent attributes have consistent units before using \code{rbindlist}

[ben-schwen]

Ty Nitish!