* Split key-value line on first colon

* Add IPv6 test to connected slave test

* add last_slow_execution_duration_seconds

* minor refactoring and a bit more test coverage

* use drone for tests

* remove old glide config file

* remove circleCi config

* Streams metrics support.

There are new key type in Redis 5 - Streams.
So, we get ability to monitor them.

* Test for stream.

* Adds tile38 metrics and parser.

* Adding replication lag metric

* Add slave_port attribute to metric "redis_connected_slave_offset"

* fixed issue#221: apply correct key name while formatting debug log

Remove `mem_fragmentation_ratio`:
* Comes from `redis_memory_used_rss_bytes / redis_memory_used_bytes`.
* Metric only provides 2 digits of accuracy compared to calculation.

Add allocator metrics.

Cleanup Prometheus naming of commands metrics names to avoid confusing
references to being a "Summary" metric, it's not.

Remove `total_system_memory`, this duplicates node_exporter metrics.

Signed-off-by: Ben Kochie <superq@gmail.com>

* Avoid using KEYS.

* Check single keys, or patterns with `SCAN`

Continues the work that @Tinche did in #149.

* Add an arg, `check-single-keys` to get size and value with `TYPE`
* Modify the behavior of `check-keys`
  * Get keys from globs with `SCAN` rather than `KEYS`
  * If a key doesn't have any glob characters in it, just add it to the
    list of things to check with `TYPE`

I added an argument rather than decide `SCAN` vs `TYPE` based on the
presence of glob characters because, technically, redis key names can
contain glob characters, according to https://redis.io/topics/data-types-intro:

    Redis keys are binary safe, this means that you can use any binary
    sequence as a key, from a string like "foo" to the content of a JPEG
    file. The empty string is also a valid key.

The only way to be able to check for key names with glob characters in
them is by being able to explicitly specify keys that you don't want to
glob for.

* add empty help text to be compatible with prometheus >= 2.3.1

* Properly decode uptime from redis
* Add uptime_in_seconds & start_time_seconds to expected metrics

start_time is a more idiomatic way to store the process uptime,
and is what is provided by the Prometheus library itself.

Since this is a breaking change, I am happy to add this as a second
metric if necessary

Allow an optional Redis Lua script to collect extra metrics.

This can be enabled with a new `-script` flag. An example is provided in
contrib.

Bump redigo to 2.0.0 for new paths