First Letter Filter

[CaffeineLab]

I found that pyspellchecker was giving me very accurate suggestions until I started using custom dictionaries with some non-standard words. For example, I had the word 'Zuro' in my local_dictionary and I was testing against 'Zury'. No matter what I did, I kept getting 'Fury' as the suggested correction, mostly because it's number of instances in the dictionary was waaaaaaay higher.... and that's fine...

So I had seen somewhere that I could increase the accuracy in my case if I just filtered out the list of candidates to those that shared the first letter as the word I was correcting.... and .... it was right.

Change isn't too horrible - but if interested I could supply it as a comment in this discussion, or do it in a PR....

[CaffeineLab]

The gist of it is as follows:

`
def candidates(....

    res = list(self.edit_distance_1(word))
    
    if self._same_start:
        res = list([i for i in res if i.startswith(word[0:1])])

`