Implement test -L, and add tests.

It's an alias for test -h, which had a bug because we were using
os.stat() and now os.lstat().

Addresses issue #58.

core/expr_eval.py

@@ -479,7 +479,7 @@ def Eval(self, node):
      arg_type = BOOL_OPS[op_id]  # could be static in the LST?
      if arg_type == OperandType.Path:
        try:
          mode = os.stat(s).st_mode
          mode = os.lstat(s).st_mode
        except OSError as e:  # Python 3: FileNotFoundError
          # TODO: Signal extra debug information?
          #self._AddErrorContext("Error from stat(%r): %s" % (s, e))
@@ -503,7 +503,7 @@ def Eval(self, node):
        if op_id == Id.BoolUnary_w:
          return os.access(s, os.W_OK)

        if op_id == Id.BoolUnary_h:
        if op_id in (Id.BoolUnary_h, Id.BoolUnary_L):
          return stat.S_ISLNK(mode)

        raise NotImplementedError(op_id)

spec/builtin-test.test.sh

@@ -197,3 +197,16 @@ chmod -w $TMP/testw_no  # remove write permission
test -w $TMP/testw_yes && echo 'yes'
test -w $TMP/testw_no || echo 'no'
# stdout-json: "yes\nno\n"

### -h and -L test for symlink
ln -s -f $TMP/zz $TMP/symlink
test -L $TMP/zz || echo no
test -h $TMP/zz || echo no
test -L $TMP/symlink && echo yes
test -h $TMP/symlink && echo yes
## STDOUT:
no
no
yes
yes
## END