Refactor Execution::Next runner flow

[ydah]

GraphQL::Execution::Runner#execute and #begin_execute currently handle multiplex analysis, query setup, step execution, result finalization, root authorization, operation directives, and root-type dispatch inline.

This PR reorganizes those responsibilities into private methods:

• Separates multiplex analysis, query setup, step execution, and result finalization.
• Separates object, abstract, list, and scalar root execution.
• Extracts root authorization, operation directive handling, and mutation setup.
• Consolidates repeated root SelectionsStep construction.
• Keeps isolated mutation step groups on the runner so setup methods can enqueue work.

The existing result ordering, tracing boundaries, directive finalization, mutation isolation, and stack-error handling are preserved.

[rmosolgo]

Could you please isolate the bug fix from the code reorganization? The bug fix is an easy include, but I doubt I want to include the code reorganization. Whenever possible, I prefer a single method because I can read it from top-to-bottom and know what's going on. When the code is divided into methods, then the reader (me?) has to jump around to find out what's really going on.

From a maintenance perspective, I prefer keeping API surface area as small as possible: private instead of public whenever possible, inline instead of extracted whenever possible, local variable instead of instance variable whenever possible, and so on.

[ydah]

@rmosolgo Thank you, that makes sense.
I've separated the changes and force-pushed this PR so it now contains only the code reorganization.