Introduce and extend Config pattern (III/2)

[sazzad16]

Resolves #1780
Resolves #1824
Resolves #2234

Closes #1787
Closes #2237

Closes #2342
Closes #2349
Closes #2352
Closes #2353
Closes #2363

[mina-asham]

Looks good, thanks for going through the comments from the prevision revision! Just a few more minor comments here and then I am happy to sign off

[sazzad16]

@gkorland

1. ShardedJedisPoolTest.shouldReturnActiveShardsWhenOneGoesOffline()
2. ShardedJedisPoolWithCompleteCredentialsTest.shouldReturnActiveShardsWhenOneGoesOffline()
3. ShardedJedisTest.testMasterSlaveShardingConsistency()
4. ShardedJedisTest.testMasterSlaveShardingConsistencyWithShardNaming()

These tests used to work because underlying connections were not always created given the combination of parameters. All the Jedis classes were suffering from this situation. The underlying connection of object representing a connection should either always or NEVER be created from constructor. In this PR I implemented the "always" approach.

I was able to keep the tests with a valid host address in (3) and (4).

But I could not make (1) and (2) to work. Even though I @Ignored first, I removed those after review. A portion of (1) and (2):

    shards.set(1, new JedisShardInfo("localhost", 1234));
    pool = new ShardedJedisPool(redisConfig, shards);
    jedis = pool.getResource();

This is more logical that pool.getResource() would fail because 1234 is an invalid port. But the tests are expecting pool.getResource() without any error. That's why these two tests won't work with this PR.

IMHO, we shouldn't care about impractical tests. Still, I can @Ignore those if you prefer.