(fix)Exposures definitions and added ability to preserve columns

[Vitor-Avila]

This PR addresses a few issues with the exposures sync:

• In the past, we were not including the single quotes to specify the model name here, which could cause dbt exposure broken references #208
• The asset titles were used as the name for exposures, which could contain incompatible characters: synced dbt exposures name need to be an identifier #207 and Exposure names contain invalid characters #177
• Charts don't have a query_context value until they are viewed in Explore, which could cause: syncing dbt exposures fail  #206

This PR also introduces a new flag to the dbt sync: --preserve-columns. By default, the sync would update existing datasets in Superset with override_columns=True, which resets the column definition (groupby and filterable configurations) and also sync columns from the dataset. This could potentially misconfigure existing columns in Superset, or bring back to the datasets columns that were intentionally excluded. This behavior will be kept by default (since the main use-case for the dbt integration is to have the dbt info as the source of truth), however if the --preserve-columns flag is included in the dbt sync command, the update dataset API calls would include override_columns=False, preserving column configurations and avoiding a column sync.

[betodealmeida]