Editorial: refactor classic and module scripts to be more alike

This makes "create a classic script" work the same as "create a module
script". Both now parse source text into the appropriate record (viz.
Script Record or Source Text Module Record), storing the result, or any
parse error, for later use by "run a X script".

As a consequence, this expands the base "script" struct to include
"record" and "parse error" fields, which were previously
module-specific.

In general, this alignment will help greatly when working on dynamic
import(), which per
https://github.com/tc39/proposal-dynamic-import/blob/23cf15a8ccef0c7174b504b33fb6b2b9be1ebecd/HTML%20Integration.md
will move more things into the shared base struct.

Note that for now, "parse error" is somewhat of a misnomer for modules,
as we propagate dependency errors into this field. This is hinted at by
leaving the name for that algorithm, "set a pre-instantiation error",
intact. We anticipate clearing up this confusion in subsequent commits,
by changing error propagation, but this was left as such in order to
keep this commit purely editorial.

Along the way, this:

* Fixes a few remaining references to module trees, as opposed to module
  graphs.
* Changes "muted errors" from a flag into a boolean.
* Modernizes the algorithm style for both create algorithms, making them
  take named parameters, stop using goto-style jumps, and use booleans
  instead of flags.
* Fixes the spec to report errors while inside the "prepare to run a
  script" / "clean up after running script" block.
* Fixes several wrong variable references within these algorithms.