fix: use virtual caret in tests

[grant-d]

What does this PR do?

It improves core test infrastructure:

While helping to fix several autocomplete bugs (eg #724) in the language server, it was difficult to understand where the "cursor" was placed in sample content within the tests.

The existing code uses the following method signature, whereby the caller has to do the math on the position parameter to establish where the caret is placed:

// Method
function parseSetup(content: string, position: number): ...

// Caller
const content = 'hello world this is great';
const completion = parseSetup(content, 10); // <---- Can you easily tell where the cursor is?
`

While this works, it is difficult to maintain since it's not easy to understand the test by just looking at the content (& without doing math)

This PR introduces the notion of a virtual caret, which is specified in the content itself.
The token |:| is chosen to be easily noticed, and not conflict with 'real' yaml

// Method
function parseSetup(content: string, position?: number): ... // position optional if caret specified

// Caller
const content = 'hello worl|:|d this is great'; // Note embedded caret token `|:|` representing cursor pos
const completion = parseSetup(content); // Zero calories
`

What issues does this PR fix or reference?

Tests are simpler to write and maintain since the cursor's position is explicit in the content.

Is it tested? How?

The changed method is back-compatible, so all existing tests continue to pass.
However, I have updated one example test of each interesting scenario:

• When the caret is a prefix, eg |:|abc
• When the caret is a postfix, eg abc|:|
• When the caret is an infix, eg ab|:|c

If this PR is approved, I will update the other tests accordingly. I updated just a few tests for now to keep the noise down and the changes easy to identify.

cc @p-spacek

[grant-d]

This is the critical change
If the position is not specified, we look inside the content to derive the position using the index of the |:| token

[grant-d]

Scenario 1: position === 0

[grant-d]

Scenario 2: position === <postfix>

[grant-d]

Scenario 3: position === <infix>

[grant-d]

Scenario 4: position === <length>

[grant-d]

Perf optimization: just call getText() once

[coveralls]

Coverage increased (+0.003%) to 82.403% when pulling a8f2c11 on p-spacek:fix/caret into a3dd4e6 on redhat-developer:main.

[gorkem]

This is so cool. First on my list for review.

[msivasubramaniaan]

LGTM

Thanks for the contribution

[grant-d]

Thanks @gorkem , @msivasubramaniaan

[p-spacek]

@grant-d love this, counting chars was an adventure :)
is it possible to add this also into autoCompletionFix.test.ts file?

[grant-d]

Yes - my plan was to introduce the concept in a bite-sized PR, then once I got buy-in, I will spread the love elsewhere