Remove OpenSSL dependency

[aido]

Hi,

This pull request removes the OpenSSL dependency. The OpenSSL RAND_bytes functionality is replaced with an implementation of Fortuna PRNG based on code from the PostgreSQL database project.
http://www.postgres.org/

A few builds are failing in Travis (see issue #91) but this is fixed in pull request #94.

With this pull request picocoin now has two implementations of AES. The ctaes.c code used by the wallet and rijndael.c code used by Fortuna. For now it was easy to just add the rijndael.c code for Fortuna but this may be changed in the future so that there is just a single AES implemetaion used by both Fortuna and the wallet.

[coveralls]

Coverage decreased (-8.6%) to 70.805% when pulling a15680e on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-6.5%) to 72.952% when pulling eaed5a7 on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-6.5%) to 72.955% when pulling 067cfd4 on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-6.4%) to 73.05% when pulling 45716c5 on aido:fortuna into 81bb7ab on jgarzik:master.

[jgarzik]

Concept ACK

Seems like valgrind and aes-util tests are failing, triggering a Travis failure.

[aido]

Hi Jeff,

Yes, these tests are failing because of issue highlighted in #91. They should be fixed in pull request #94.

I have also written a test to check just how random the new PRNG actually is. It performs the same tests as the ent utility (https://www.fourmilab.ch/random/). However, I am finding that due to the nature of PRNGs this test randomly fails so I have made it an optional test by setting FORCE_PRNG_VERF=1.

The test currently gives the following output, which looks positive:

Entropy:
========
Entropy = 8.000000 bits per byte.

Optimum compression would reduce the size
of this 536870912 byte input by 0 percent.

Chi-square Test:
================
Chi square distribution for 536870912 samples is 264.59, and randomly
would exceed this value 32.68 percent of the times.

Arithmetic Mean:
================
Arithmetic mean value of data bytes is 127.4962 (127.5 = random).

Monte Carlo Value for Pi:
=========================
Monte Carlo value for Pi is 3.141875994 (error 0.01 percent).

Serial Correlation Coefficient:
===============================
Serial correlation coefficient is 0.000044 (totally uncorrelated = 0.0).

See https://www.fourmilab.ch/random/ for detailed description of output.
Repeatedly running the test does not always give as favorable results as above.

Also, the PRNG can easily be expanded with other sources of entropy in the future.

[coveralls]

Coverage decreased (-6.2%) to 73.217% when pulling 310456f on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-7.9%) to 71.508% when pulling c22ecf2 on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-7.9%) to 71.508% when pulling c22ecf2 on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-7.5%) to 71.916% when pulling bd13d53 on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-7.5%) to 71.916% when pulling bd13d53 on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-7.5%) to 71.916% when pulling f848b4c on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-7.5%) to 71.923% when pulling c7bb0a7 on aido:fortuna into 81bb7ab on jgarzik:master.

[coveralls]

Coverage decreased (-7.4%) to 72.013% when pulling c7bb0a7 on aido:fortuna into 81bb7ab on jgarzik:master.