Splitting arrays/objects over code blocks breaks syntax highlighting

Description

In literate mode, splitting an array or object over more than one code block breaks syntax highlighting as shown below.

Steps to Reproduce

Create a new file.
Save it with the .litcoffee extension.

Enter the code below.


Some documentation, highlighted correctly.

myArray = [
    'item1'

Some documentation, highlighted as code. :(

    'item2'
    'item3'
]

Things are still not right.


**Expected behavior:** Text without indentation should always be treated and highlighted as Markdown, never mistaken for Coffeescript code.  GitHub highlights it incorrectly as well, as you can see immediately above.

**Actual behavior:** The following screenshot shows that the syntax highlighter has begun to treat unindented code as if it were Coffeescript code.

<img width="478" alt="screen shot 2017-10-25 at 10 38 58 am" src="https://user-images.githubusercontent.com/1316622/32004934-019c58dc-b971-11e7-93bd-10ef3bcad1fc.png">

The same behavior occurs with objects instead of arrays, as shown below.

```litcoffee
Some documentation, highlighted correctly.

    myObject = {
        'item1' : 1

Some documentation, highlighted as code. :(

        'item2' : 2
        'item3' : 3
    }

Things are still not right.

According to this page, it seems like the GitHub syntax highlighting is actually handled by this package (atom/language-coffee-script), which is a little surprising to me, since there have been differences in the past on how the two behave, but perhaps this is a recent change? If need be, I'm glad to file an issue in the appropriate other repo so that the issue is fixed on GitHub also.

Reproduces how often: 100%

Versions

Atom : 1.21.0

Electron: 1.6.9

Chrome : 56.0.2924.87

Node : 7.4.0

Mac OS X 10.11.6

atom / language-coffee-script