Support configurable bucket ranges for bulk_process_time Histogram

[ful09003]

Greetings,

First time contributor, please let me know if this is an acceptable PR!

To address #45, I am proposing a set of changes which amount to:

• Allow for float64 inputs from Firebolt config
• Add 3x new configuration options on the elasticsearch node type, related to Histogram bucket configuration, min, max, and count (matching the inputs of prometheus.ExponentialBuckets/prometheus.ExponentialBucketsRange)
• Begin passing in user-supplied options, or what should be safe defaults. For defaults, I opted to match the existing Firebolt choices as closely as possible:
  • 0.01 for a bucket min
  • 2 * (bulk-index-timeout-ms / 1000) for a bucket max. I'm open to changing this, however it seems appropriate to clamp the maximum this way to reduce the chances of observations being lumped into +Inf which is not as helpful when often timeouts are 🤏 just barely (by a few seconds, at most) over bulk-index-timeout-ms
  • 8 buckets

Using the defaults above, and assuming a 5000ms bulk-index-timeout-ms in unit testing, this PR results in the following buckets on my machine: 0.01, 0.026826957952797256, 0.07196856730011518, 0.19306977288832497, 0.5179474679231209, 1.3894954943731366, 3.7275937203149367, 9.999999999999991

[ful09003]

Mmmm, in hindsight of course unit testing assumptions from one platform would not work (and we see that here, with the buckets generated in unit testing differing very slightly from the ones I had, locally). I'll think about how to address that and come back once ready!

[jnadler]

thank you for your contribution!