Handle invalid timezone values

[tomchy]

Possible fix for #15
Returned timezone value is not a valid BCD number (0x2C). The timezone value should be close to +7 (+28 quarters, 0x28 in BCD). Currently I can see four interpretations:

• not masked, badly positioned sign bit: 0x2C & 0x77 = 0x24 (+6 h)
• badly positioned sign bit: 0x2C = -0x24 (-6 h)
• not reorganized bytes: 0x2C -> 0xC2 = -0x42 (-10.5 h)
• overloaded BCD value: 0x2C + 0x6 = 0x32 (+8h)

According to the fact that the SMS was send from Jakarta (+7), negative values are (most probably) invalid (option 2 and 3).
Indonesia has timezones +7, +8 and +9, so 4th option is more probable that the 1st one (why local message should leave the country?).

This PR changes timezone calculation using 4th option and redefines unit test, which has problematic status PDU.

[coveralls]

Coverage decreased (-0.01%) to 90.021% when pulling 2849814 on tomchy:pr/pdu-timezone-fix into 459159f on babca:master.

[babca]

You've done some solid detective work here.