`precondition()` should be an error thrown

[herzi]

Describe the issue

PostgresRow.decode(_:context:file:line) triggers a precondition failure when the number of column types does not match the number of returned column

Vapor version

vapor=n/a postgres-nio=1.33.0

Operating system and version

macOS 26.5.1

Swift version

swift-driver version: 1.148.6 Apple Swift version 6.3.3 (swiftlang-6.3.3.1.3 clang-2100.1.1.101)

Steps to reproduce

Use a code snippet like this to iterate a mismatch query/decode:

let client: PostgresClient // populate however you prefer
try await client.withConnection { connection in
  // note how the query just returns one value
  for try await row in try await connection.query("SELECT now() as time", logger: logger) {
    // note how we try to decode more than one column
    try row.decode((Date, String).self) 
  }
}

Expected outcomes:

1. Code compiles
2. Code runs and throws an error

Outcome

Observed outcomes:

1. Code compiles
2. Code crashes with PostgresNIO/VariadicGenerics.swift:56: Precondition failed (which also isn't a very good error message for this thing happening)

Additional notes

While this is clearly a programming error (on my side), I think it would be more consistently handled with the rest of the framework's behavior. A programming error is also an invalid SQL query, however that will throw PSQLError, so that other tasks in a Swift-based server can continue.

More importantly, in a multi-service environment, throwing an error will at least provide a way to shut down other services cleanly compared to a hard crash that might lose data in-flight.

NOTE: If maintainers agree that this is a useful improvement, I will happily work on a PR to implement this.

[gwynne]

I definitely concur that this is a useful improvement! This absolutely should not cause a hard crash. It is technically a programming error, but such a mismatch can also be caused by changes on the database side over which the server has no control, such as a stored procedure's output having a different format after a schema update.

[herzi]

Thanks a lot for the feedback, @gwynne. I raised PR #650 to address this and would love to get some feedback.