docs: Add custom data quickstart guide and fix instruct dataset examples

- Add comprehensive custom_data_quickstart.rst guide
- Fix missing _component_ fields in instruct_datasets.rst examples
- Complete implementation for issues #2215 and #2221

Author: JonSnow1807

docs/source/basics/instruct_datasets.rst

@@ -66,6 +66,7 @@ Here is an example of an instruct dataset to fine-tune for a grammar correction
       max_seq_len: 8192
 
     dataset:
+      _component_: torchtune.datasets.instruct_dataset
       source: csv
       data_files: data/my_data.csv
       split: train