Fixes bug splitting the statements into commands

[ncordon]

What

When splitting the commands, we try to parse the input as cypher or console commands, etc. If all of that fails, we default to taking the full query if a cypher statement could not be parsed, but we were taking a stop token that could possibly be null. This PR just defaults to taking the full stream when that token is null:

    // stop here could be null
    const statement = inputstream.getText(start.start, stop.stop);
    return { type: 'cypher', statement, start: start, stop: stop };

Why

Because we were blowing up in multiline comments (where the parsing tree would be empty):

[changeset-bot]

🦋 Changeset detected

Latest commit: 5ccba29

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 5 packages

Name	Type
@neo4j-cypher/language-support	Patch
@neo4j-cypher/language-server	Patch
@neo4j-cypher/react-codemirror-playground	Patch
@neo4j-cypher/react-codemirror	Patch
@neo4j-cypher/schema-poller	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[anderson4j]

Mostly LGTM, but would we not get the same issue with unfinished strings/escaped sequences? (If so, we might as well add tests for these too)

And if my summary of the issue (see comment) is correct, would we encounter it for single line comments?

[anderson4j]

Double checking my understanding - I would explain like:
"When ANTLR can't finish the parsing (but can tokenize) because a comment/string/escaped sequence is unfinished we get here with stop===null"

[ncordon]

The problem if we just a comment is that's not passed to the parser, so the CST is empty, therefore because the parser expects something, it errors and cannot find a stop. Strings and backticked elements are not on the hidden channel, so they are passed to the parser

[anderson4j]

aha, your response jogged my memory - comments are removed just like blank space when tokenizing, so what would be the stop-token when parsing to command doesnt exist

Think changeset-message is off then, the issue is not with "non-finished multiline comments" (that's more like the bug in the other PR right) but rather "queries ending in a comment"

approving so you can merge once that's adjusted