English tokeniser skips negative numbers

[AndrioCelos]

If the text input includes negative numbers, such as -1, those may be skipped over. For example, g2p('-1')[0] returns the empty string.

The expectation was that it would be read as a negative number, such as /mˈInəs wˈʌn/.

If trf is set to True, negative numbers are pronounced correctly if other words precede them, but the above example still results in the empty string.

[hexgrad]

This is caused by the spacy POS tags (see glossary) being incorrect for -1. Thinking of a fix for this, thanks for the issue.