String length operators now counts code points, not bytes (#145)

The length operator ${#s} returns the number of code points now instead
of the number of bytes. Also added a test for this. Addresses #142.

* Change hard tab to soft tab

core/word_eval.py

@@ -298,7 +298,9 @@ def _ApplyPrefixOp(self, val, op_id):

    if op_id == Id.VSub_Pound:  # LENGTH
      if val.tag == value_e.Str:
        length = len(val.s)
        unicode_val = val.s.decode('utf-8')
        length = len(unicode_val)
        # length = len(val.s)
      elif val.tag == value_e.StrArray:
        # There can be empty placeholder values in the array.
        length = sum(1 for s in val.strs if s is not None)

spec/var-op-other.test.sh

@@ -9,6 +9,13 @@ v=foo
echo ${#v}
# stdout: 3

### Unicode string length (UTF-8)
v=$'_\u03bc_'
echo ${#v}
## stdout: 3
## BUG dash stdout: 9
## BUG mksh stdout: 4

### Length of undefined variable
echo ${#undef}
# stdout: 0
@@ -158,4 +165,3 @@ echo ${foo:1:3}
# BUG mksh stdout: -μ
# N-I dash status: 2
# N-I dash stdout-json: ""