write semantics

We currently don't have documented or even well defined write semantics for the integration:
```
def write(
self,
df: pd.DataFrame,
table_name: str,
pks: List[str] = None,
as_key: str = None,
)
```

This does not specify where a write will create, update, or replace a table. Furthermore it doesn't handle update-on-duplicate.

We need to figure out what kind of semantics we want to expose and pipe them through to the UX.

Start from the write(...) entry point shown in the issue and determine the intended create, update, replace, and update-on-duplicate behavior for tables. Trace how those semantics should be exposed through the UX; done means the behaviors are defined, documented, and consistently wired through the integration.

Written by the indexing model from the issue text.