Fixing requirements object not iterable on bad schema.

[nicholasks]

This will avoid raise "'NoneType' object is not iterable" error when user has a bad schema regarding requirements.txt on .pyup.yml file.

Example schema that was triggering the error:

# Specify requirement files by hand, default is empty
# default: empty
# allowed: list
requirements:
#  - requirements/staging.txt:

Should we warn users about the bad schema? And how about the documentation (.pyup.yml example file), maybe the "default empty" is leading users to do that, and I think is in certain way controversial, since below says "allowed: list".

[codecov]

Codecov Report

Merging #393 into master will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master     #393   +/-   ##
=======================================
  Coverage   94.20%   94.21%           
=======================================
  Files          11       11           
  Lines        1157     1159    +2     
=======================================
+ Hits         1090     1092    +2     
  Misses         67       67           

Impacted Files	Coverage Δ
pyup/config.py	98.91% <100.00%> (+0.02%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 6fb8638...deab8bd. Read the comment docs.

[rafaelpivato]

We had something like this happening before and fixed with #395. We certainly need better schema validation for our configuration file. Do you have suggestions?

If you don't want to bring that with this PR, I would just urge you to consider, besides an empty None value for requirements, what could happen if someone adds unwelcome or unwanted types, like hashes instead of lists, integers instead of strings and so on.