fix(reader): handle header-only results

[mharrisb1]

Closes #62

Issue was that single cell results are the same as header-only results. For header-only results we returned bind data too early without setting column name and types so when bind function was called by DuckDB it failed due to empty schema error.

This makes the following changes to avoid this:

• Catch empty results and throw custom exception to tell user the exact cause of error
• Avoid segfault when trying to access first row at index 1 when header-only results

The following tests were added to catch these issues in the future

• Read sheet with single cell range (w/ headers)
• Read sheet with single cell range (w/o headers)
• Read header-only sheet
• Read empty sheet

[archiewood]

Played around with this, big improvement.

The only error that could be clearer is this one

from read_gsheet('https://docs.google.com/spreadsheets/d/11QdEasMWbETbFVxry-SsD8jVcdYIT1zBQszcF84MdE8/edit?gid=0#gid=0', range='d4');

Invalid Input Error:
Sheet Sheet1 is empty

In this case, Sheet1 is not empty, but the selected range is empty. I think users will be smart enough to work this out though, and feels like a low % edge case.

[mharrisb1]

I'm about to work on ranged writes and can make this change then. Do you have a better alternative?

Maybe something about values being empty instead of sheet? That's really what this would be, the API fetch results values field is empty

[archiewood]

How about something like this if the range is empty?

Error:
Range Sheet1!D1:D2 is empty

Else - we could do something less specific, which we could throw irrespective of whether the user has specified a range or not

Error:
The specified range in Sheet1 is empty