Add sink timestamp section into file names

[chuwy]

Currently file names are generated here: https://github.com/snowplow/snowplow-s3-loader/blob/master/src/main/scala/com.snowplowanalytics.s3/loader/KinesisS3Emitter.scala#L150 with only year/month/day and Kinesis sequence number, which are mostly useless if we want to reply some set of data from S3.

/cc @istreeter

[colmsnowplow]

There is another purpose to this requirement, and an important one. If the path doesn't follow a particular convention, then it is not possible to partition the data and limit queries via Athena.

Ideally the S3 loader will be able to create files in a manner which allows one to easily create partitioned athena tables, and only load certain partitions. (the old format run= convention served this purpose well).

Edit for clarity: The convention to follow is key=value in the name, eg. run=2020-04-23. I'm unsure at present whether this must be a directory name, or this convention in the filename is sufficient.

[benjben]

We could also use this parameter to partition data by date.

[benjben]

2020-12-22-49613548169053493378838656625866917741098839399456571394-49613548169053493378838657029012246511267725570343960578.gz becomes 2020-12-22-125000-49613548169053493378838656625866917741098839399456571394-49613548169053493378838657029012246511267725570343960578.gz where 125000 is for 12:50:00