feat(c/driver/postgresql): Support queries that bind parameters and return a result

[paleolimbot]

This PR allows statements with a bind stream to return a result. This is not trivial because it requires concatenating results from multiple rows in a bind stream into one output stream. The existing bind stream had been wired up to pull and loop in one big function, but we need something with that knows which array/row it's on and can "bind and execute" the next one. I think I did this successfully without modifying the existing binding behaviour (which could maybe use the copy writer at some point in the future for better type support).

The behaviour of the "result reader" now incorporates the bind stream:

• When the result reader is being created/initialized, it executes eagerly (the query if there are no bind parameters, and the query for all the bind parameters that return zero-row results or until the bind stream is exhausted otherwise). If there is no out array stream, it will exhaust the bind stream before returning. This mirrors what currently happens and I think is more of what a user would expect; however, it does mean that a user would have to remember to fully consume the output array stream to ensure that the full bind stream was exhausted (if they specified an output stream in the first place).
• On each call to GetNext() of the result, we check for an existing PGresult to convert, and try to regenerate it if it's absent (== already been converted to an array and returned) using a new value from the bind stream. When pulling from the bind stream we pull everything we can until we get something with at least one row.

Much existing behaviour (update, create table, insert, delete) flows through the first bullet, and I believe this has some more correct behaviour for the case where an insert or update or delete updates a number of rows that isn't one and/or there is more than one value in the bind stream.

library(adbcdrivermanager)

con <- adbc_database_init(
  adbcpostgresql::adbcpostgresql(), 
  uri = "postgresql://localhost:5432/postgres?user=postgres&password=password"
) |> 
  adbc_connection_init()

read_adbc(con, "SELECT $1 as x", bind = data.frame(x = c("a", NA, "b", "c"))) |> 
  as.data.frame()
#>      x
#> 1    a
#> 2 <NA>
#> 3    b
#> 4    c

execute_adbc(con, "DROP TABLE IF EXISTS integers")
write_adbc(data.frame(x = 1:10), con, "integers")

read_adbc(con, "SELECT * from integers where x > $1", bind = data.frame(x = 8:10)) |> 
  as.data.frame()
#>    x
#> 1  9
#> 2 10
#> 3 10

^{Created on 2024-08-13 with reprex v2.0.2}

Closes #2024.

[lidavidm]

nit: we either need to exhaust the bind stream, or ...?