Add float list features in ItemSequence

- New class FloatFeatures defined in `pycrfsuite._float_features`
- Useful for adding word embedding features
- Wrap the word embedding list of float values with FloatFeatures class
- The api supports the existing extraction of features from nested
  dicts.
- Added test case in `tests/test_itemsequence.py`

Example usage:
```
import pycrfsuite
from pycrfsuite._float_features import FloatFeatures as FF
seq = pycrfsuite.ItemSequence([
  {"w2v": FF([1., 2., 3.])},
  {"w2v": FF([-1., 5, 4.])}
])
```

Author: napsternxg

pycrfsuite/_float_features.py

@@ -0,0 +1,15 @@
+from __future__ import print_function, absolute_import
+
+
+class FloatFeatures(object):
+    def __init__(self, values):
+        if not isinstance(values, list):
+            raise ValueError("Values should be an instance of list()")
+        try:
+            self.values = [float(v) for v in values]
+        except ValueError:
+            print("All elements in values should be castable to type float.", file=sys.stderr)
+            raise
+        except:
+            print("Unexpected error:", sys.exc_info()[0])
+            raise