feat: Omit table name when only one ident in SELECT

[MarinPostma]

fixes #1081

[max-sixty]

This is good; I think providing more concise SQL is important and this contributes for small queries.

I was trying to think if it would handle queries with multiple stages, like:

from employees
take 5
group [name] (
  aggregate [sum salary]
)

On main this compiles to:

WITH table_0 AS (
  SELECT
    employees.*   # would this be `*`?
  FROM
    employees
  LIMIT
    5
)
SELECT
  name,
  SUM(salary)
FROM
  table_0
GROUP BY
  name

Unfortunately semantic can't compile that yet, so I couldn't try it out. But I think it will, since this is on the sql_of_atomic_table, and so counts the tables in the CTE rather than the whole query.

If it doesn't yet, no stress, though maybe we add a comment in the code.

I'll let @aljazerzen take a glance too.

Thanks @MarinPostma !

[MarinPostma]

@max-sixty I think it does. This:

from employee
take 1..10
sort name
take 1..5

compiles to:

WITH table_0 AS (
  SELECT
    * AS _expr_0
  FROM
    employee
  LIMIT
    10
)
SELECT
  _expr_0
FROM
  table_0
ORDER BY
  name
LIMIT
  5

#1096 fixes the aliasing issue

[max-sixty]

Excellent!

Feel free to merge if you're confident overall (easy to revert if @aljazerzen spots something)

[aljazerzen]

This is great! I dont even have any suggestions!