Add sequence truncation support for long sequences

**Is your feature request related to a problem? Please describe.**
Currently, the dataset formatting utilities (ChatDataset, ColumnMappedTextInstructionDataset) do not truncate sequences that exceed certain length. `seq_length` is only used for padding, not truncation.

**Describe the solution you'd like**
Truncation support so that user can specify max sequencelength. 

