Refix fenced code blocks w/o preceding blank lines

Change approach at fixing #45: don't patch input markdown at preprocess
pass, instead improve special case detection when parsing paragraphs.

Leave the fenced code block detection in the preprocess pass though,
it's been put to another use since then, to suppress tab expansion
inside code blocks.

block.go

@@ -1342,6 +1342,14 @@ func (p *parser) paragraph(out *bytes.Buffer, data []byte) int {
 			return i
 		}
 
+		// if there's a fenced code block, paragraph is over
+		if p.flags&EXTENSION_FENCED_CODE != 0 {
+			if p.fencedCode(out, current, false) > 0 {
+				p.renderParagraph(out, data[:i])
+				return i
+			}
+		}
+
 		// if there's a definition list item, prev line is a definition term
 		if p.flags&EXTENSION_DEFINITION_LISTS != 0 {
 			if p.dliPrefix(current) != 0 {

markdown.go

@@ -393,7 +393,6 @@ func firstPass(p *parser, input []byte) []byte {
 		tabSize = TAB_SIZE_EIGHT
 	}
 	beg, end := 0, 0
-	lastLineWasBlank := false
 	lastFencedCodeBlockEnd := 0
 	for beg < len(input) { // iterate over lines
 		if end = isReference(p, input[beg:], tabSize); end > 0 {
@@ -405,16 +404,13 @@ func firstPass(p *parser, input []byte) []byte {
 			}
 
 			if p.flags&EXTENSION_FENCED_CODE != 0 {
-				// when last line was none blank and a fenced code block comes after
+				// track fenced code block boundaries to suppress tab expansion
+				// inside them:
 				if beg >= lastFencedCodeBlockEnd {
 					if i := p.fencedCode(&out, input[beg:], false); i > 0 {
-						if !lastLineWasBlank {
-							out.WriteByte('\n') // need to inject additional linebreak
-						}
 						lastFencedCodeBlockEnd = beg + i
 					}
 				}
-				lastLineWasBlank = end == beg
 			}
 
 			// add the line body if present