Parsing of whitespaces between consecutive encoded words in Subject header

This might be related to #93.

**Example**
```
Subject: =?utf-8?B?55Sw?=
  =?utf-8?B?55Sw?=
```

Expected: `田田`
Actual: `田 田` (unwanted space between them)

Since the space results from folding it should not be interpreted as an actual space.

This causes many of my longer message subjects getting sprinkled with spaces in (seemingly) random positions.