cue/token: relative positions apply before, not after, a token

Maximilian Gaß correctly points out that the cue/format package
follows relative positions before printing a token, and he's right.
I also confirmed that the token scanner also builds relative positions
based on the kind of whitespace it found leading up to each token.

While here, fix a typo.

Signed-off-by: Daniel Martí <[email protected]>
Change-Id: I2eae7f628a4e585e966e029d9d78c7fe3be89539
Dispatch-Trailer: {"type":"trybot","CL":1186143,"patchset":1,"ref":"refs/changes/43/1186143/1","targetBranch":"master"}

cue/token/position.go

@@ -123,16 +123,16 @@ const (
 	// not rendered at all.
 	Elided
 
-	// NoSpace indicates there is no whitespace after this token.
+	// NoSpace indicates there is no whitespace before this token.
 	NoSpace
 
-	// Blank means there is horizontal space after this token.
+	// Blank means there is horizontal space before this token.
 	Blank
 
-	// Newline means there is a single newline after this token.
+	// Newline means there is a single newline before this token.
 	Newline
 
-	// NewSection means there are two or more newlines after this token.
+	// NewSection means there are two or more newlines before this token.
 	NewSection
 
 	relMask  = 0xf
@@ -175,7 +175,7 @@ func (p Pos) IsValid() bool {
 }
 
 // IsNewline reports whether the relative information suggests this node should
-// be printed on a new lien.
+// be printed on a new line.
 func (p Pos) IsNewline() bool {
 	return p.RelPos() >= Newline
 }