Used a cheaper type check in toMap function, to improve performance

[mlewand]

Suggested merge commit message (convention)

Other: Improved toMap method performance. This results with improved editor data processing speed. Closes ckeditor/ckeditor5#5854.

---

Additional information

lodash' isPlainObject is a pretty performance expensive function, so should be avoided in places that are called frequently. This PR reverses the condition so a cheaper and more common condition is used first.

Alternatively we could create our own, more optimized isPlainObject implementation, but since our code uses iterables for the most part here, it's enough to just reverse the condition check.

You can use manual tests added in ckeditor/ckeditor5#5880 to have use performance tests.

There's a sub pr for this: ckeditor/ckeditor5-engine#1822.

Overview

tl;dr

test case	total time (ms/%)	PR total time (ms/%)	total click handler time	PR total click handler time
rich editor setData small content	7.6ms / 5.7%	0.5ms / 0.4%	131.92ms	129.39ms (Reduced by ~1.92%)
rich editor setData medium content	805.4ms / 16.2%	21.4ms / 0.5%	4.99s	4.02s (reduced by ~19.4%)
performance/setdata full websites (styled)	632.2ms / 13.7%	20.7ms / 0.5%	4.23s	3.55s (reduced by ~%16)

RichEditor (short semantic setData)

Before:

After:

RichEditor (medium semantic setData)

Before:

After:

setData (full websites)

Before:

After:

[coveralls]

Coverage remained the same at 100.0% when pulling b1bf335 on i/5854 into b57fc3f on master.

[oleq]

Guys, make sure these PRs are backed by noticeable performance gain confirmed in Chrome Dev Tools. For instance, do some heavy editor startup check before and after.

[jodator]

LGMT 👍 - I have added a cross-link to have this method be more discoverable.