Number protection

[PyJay]

[gabrielcnr]

Thanks for submitting this Pull Request, @PyJay . Can you please describe briefly what is the issue you are trying to solve here, or the new functionality you're trying to add?

[PyJay]

@gabrielcnr This merged pull request bypassed using a lambda: None as a callable and 'cheated' by using a dict instead to make tests in Python 2.7 pass

Today I realised that this also had an issue as len(dict) was raising a TypeError which was not being caught at the top function (ls) level. PR9 intends to fix this.

I was not sure whether Python 2.7 should be continued to be supported, however, I got curious about the initial problem and looked into it.

Turns out the lambda: None attribute breaks down by iter_ls to a 1L in python 2.7 (I haven't looked into the exact stacktrace that gets it to this point - I just checked at the 100th call of iter_ls)

>>> one = 1L
>>> one.denominator
1L
>>> id(one) == id(one.denominator)
False

As you can see, it looks like 1L is not interned and this is causing an infinite loop as the denominator (and numerator) of 1L is also 1L. This PR attempts to fix this using special treatment for numbers...