allow %d and %m to take either 1 or 2 characters, e.g. 03/04/199… (#160)

* First shot, but still no solution to #123

The problem seems that dates like 1/5/1999 are not read with
format "%m/%d/%Y", because 01/05/1999 is expected. I think
that if you read ?strptime well, this is correct, but then base
R works differently and accepts single digit days and months. So,
I think the preferred behavior is to accept them.

This PR is messy and full with print statements. I believe that the
line marked with "need to set back the parser iterator/pointer one
character here?" should be replaced with setting back the parser
pointer one step, as we have a fail read here which does NOT raise
an error. The error then happens in next step, when the "/" (or other
separator) has been absorbed, I suspect. Happy to revisit this when
I know how the parser works.

* allow %d and %m to be 1 or 2 characters long; fix #123

* tidy

* add NEWS entry

Fixes #123
Fixes #170

NEWS.md

@@ -1,5 +1,7 @@
 # vroom (development version)
 
+* `col_date` now parses single digit month and day (@edzer, #123, #170)
+
 * `vroom_fwf()` now handles files with dos newlines properly.
 
 * Added benchmarks with _wide_ data for both numeric and character data (#87, @R3myG)

src/DateTimeParser.h

@@ -214,7 +214,7 @@ class DateTimeParser {
         year_ += (year_ < 69) ? 2000 : 1900;
         break;
       case 'm': // month
-        if (!consumeInteger1(2, &mon_, false))
+        if (!consumeInteger1length1_or_2(2, &mon_, false))
           return false;
         break;
       case 'b': // abbreviated month name
@@ -226,7 +226,7 @@ class DateTimeParser {
           return false;
         break;
       case 'd': // day
-        if (!consumeInteger1(2, &day_, false))
+        if (!consumeInteger1length1_or_2(2, &day_, false))
           return false;
         break;
       case 'a': // abbreviated day of week
@@ -434,6 +434,24 @@ class DateTimeParser {
     return true;
   }
 
+  // Integer indexed from 1 (i.e. month and date) which can take 1 or 2 positions
+  inline bool consumeInteger1length1_or_2(int n, int* pOut, bool exact = true) {
+    int out1, out2;
+    if (!consumeInteger(1, &out1, true))
+      return false;
+    else {
+      if (consumeInteger(1, &out2, true))
+        *pOut = 10 * out1 + out2;
+      else {
+        *pOut = out1;
+        dateItr_--; // unconsume the last read non-integer char
+      }
+    }
+
+    (*pOut)--;
+    return true;
+  }
+
   // Integer indexed from 1 with optional space
   inline bool consumeInteger1WithSpace(int n, int* pOut) {
     if (consumeThisChar(' '))