Improve performance of package comment regex

[kalafut]

By anchoring to the start of the document (since that's the only place
a package comment will appear), the number of false matches is greatly
reduced in some files.

Fixes #1747

[arp242]

A package comment is not always at the start of the document; it can be preceded with one or more other comment blocks.

For example this file will no longer work correct:

// +build foo

/*
Copyright ...
*/

// Package main should be folded
// fold this.
package main

func main() {
}

I'm not sure if there is a way to make this significantly faster; maybe using \%^.* is faster, or maybe not.

Screenshots

Master:

This PR:

[kalafut]

OK, I see what you mean. Let me see if I can adjust this.

[kalafut]

I played around with a number of regex that mostly handled these cases, but they became uncomfortably complex. Taking a step back, the real performance problems stem from the /* ... */ matching, since they can be started with any /* and end up searching the whole rest of the file. So a completely different approach to this PR is simply to replace:

start=/\v\/\*.*\n(.*\n)*\s*\*\/\npackage/ with
start=/\v^\s*\/\*.*\n(.*\n)*\s*\*\/\npackage/

This start of line anchor helps a lot because it prevents quoted /* from starting new searches, and is (I think?) still valid for package comments.

Note that if there are block (/* .. */) comments (like a license) and then a block commented package comment, the license+package get folded together, like you were just mentioning. But that behavior is present even in the released version, and line comment style is more common anyways.

If you like this approach I'll update that PR with it.

[arp242]

Seems like a good solution 👍 Thanks!