Fix Return Type For to_pem

[munagekar]

>>> import ecdsa
>>> private_key = ecdsa.SigningKey.generate(curve=ecdsa.NIST256p)
>>> pub_key_pem_bytes = private_key.get_verifying_key().to_pem()
>>> type(pub_key_pem_bytes)
<class 'bytes'>

[coveralls]

Coverage decreased (-0.05%) to 96.989% when pulling ac431c5 on munagekar:master into e24c027 on warner:master.

[tomato42]

yes, the documentation indeed is incorrect

that being said, the PEM encoding should be str on python3 as both the headers as well as the base64 encoding inside are supposed to use human-readable characters (i.e. when PEM specifies "A" it means Unicode 0x0041 LATIN CAPITAL LETTER A, not a single byte with value 65)

fixing it would break API so it's not something I'd like to do, so I guess we need to live with it, but I'd prefer to see a warning in the doc string that the PEM byte string returned is already encoded to US-ASCII so it needs to be re-encoded if the system or protocol in which it is used in is not compatible with US-ASCII (e.g. uses UTF-16 or EBCDIC)

[tomato42]

please add a warning about the bytes being returned

[munagekar]

Thank you for the review. I have added a warning.

[tomato42]

thanks!