Optimize traversing the DOM when analyzing text content

Previously, this was an often-repeated construct in readability implementation:

    charCount(ps.getInnerText(node, true)) == 0

What this would do is:

- Call `dom.TextContent(node)` to append the contents of all individual text nodes together;
- Pass the result through `strings.TrimSpace`;
- Pass the result through the NormalizeSpaces regex which squashes consecutive runs of whitespace;
- Count the Unicode runes of the result;
- Finally, if the count is zero, the element would be considered "empty".

The above is an example of an incredibly costly operation that could be done much more efficiently, for example: walk the DOM subtree until the first non-space character is found, then bail out and conclude that the element has content. This barely needs any memory allocations, and is the approach taken in this PR to address a variety of counting or detecting tasks that share a similar purpose.

Benchmark before vs. after for processing a large HTML document reveals significant saving in memory allocations:

variant | times | ns/op      | Bytes/op   | allocs/op
--------|-------|------------|------------|----------
before  | 30    | 38,986,203 | 59,623,683 | 199,876
after   | 36    | 31,910,769 | 11,449,004 | 119,810

Author: mislav

internal/re2go/grab-article.go

@@ -37,25 +37,3 @@ func MaybeItsACandidate(input string) bool {
 		*/
 	}
 }
-
-// Commas as used in Latin, Sindhi, Chinese and various other scripts.
-// see: https://en.wikipedia.org/wiki/Comma#Comma_variants
-// Original pattern: \u002C|\u060C|\uFE50|\uFE10|\uFE11|\u2E41|\u2E34|\u2E32|\uFF0C
-func CountCommas(input string) int {
-	var count int
-	var cursor, marker int
-	input += string(rune(0)) // add terminating null
-	limit := len(input) - 1  // limit points at the terminating null
-	_ = marker
-
-	for { /*!use:re2c:base_template
-		re2c:case-insensitive = 1;
-
-		commas = [\u002C\u060C\uFE50\uFE10\uFE11\u2E41\u2E34\u2E32\uFF0C];
-
-		{commas} { count++; continue }
-		*        { continue }
-		$        { return count }
-		*/
-	}
-}

internal/re2go/grab-article.re

@@ -152,11 +152,6 @@ func Test_MaybeItsACandidate(t *testing.T) {
 	assert.False(t, MaybeItsACandidate(`<p class="text">Paragraph text</p>`))
 }
 
-func Test_CountCommas(t *testing.T) {
-	assert.Equal(t, 3, CountCommas("my,name,is,john"))
-	assert.Equal(t, 9, CountCommas("now,its،a mixed﹐commas︐from︑various⹁place⸴and⸲country，"))
-}
-
 func Test_NormalizeSpaces(t *testing.T) {
 	assert.Equal(t, "some sentence", NormalizeSpaces("some   sentence"))
 	assert.Equal(t, "with tabs", NormalizeSpaces("with \t \ttabs"))