Data tab should reflect changes in the columns of the data file

**System Information (please complete the following information):**
latest versions as of 14.7.2021, windows 10

**Describe the bug**
If the CSV file has changed, by adding new or removing old columns, it is not reflected in Data tab. If trying to retrain with a changed CSV file, it may fail because it tries to read wrong column, or worse, it may not fail but uses wrong columns from CSV to each feature.

**To Reproduce**
1. Train a model with any CSV file
2. Edit CSV file and add new columns (or edit .mbconfig JSON to point to a new CSV file)
3. Come back and observer columns shown in Data view. Note that the new columns are not visible.

**Expected behavior**
Data tab seems to re-read the file (because it is slow to open), it also should update the columns if any changes.

**Additional context**
Low priority. Workaround is to remove CSV file and then re-select it in the Data tab.