Metrics collecting and reporting support

[karalabe]

The goal of this PR is that - similar to logs - we should be able to add metrics collection to any part of the code without fancy constructs (counter variables, exporting them through public interfaces, adding them to the APIs, consoles, RCP/IPC, etc), and have the just work, be queryable and visualizable.

To that extent, this PR currently adds support for two metric types:

• Meters: Event counters that can provide rate averages for the past 1, 5, 15 mins, and totals.
• Timer: Timed meters, where each event has a duration associated and the last 1024 can be queried.

To create and use a meter and/or timer:

meter := metrics.GetOrRegisterMeter("my/fancy/meter", metrics.DefaultRegistry)
meter.Mark(125) // Register 125 events (e.x. 125 bytes written).

timer := metrics.GetOrRegisterTimer("my/fancy/timer", metrics.DefaultRegistry)
timer.Update(time.Second) // Register an event that took a second (e.x. block import)

These are automatically registered into a central metrics registry and can be queried from the javascript console through debug.metrics() for a user friendly pre-formatted output:

> debug.metrics().p2p.InboundTraffic
{
  Avg01Min: '169.12K (2.82K/s)',
  Avg05Min: '1.92M (6.42K/s)',
  Avg15Min: '3.57M (3.96K/s)',
  Total: '5.83M (2.97K/s)'
}
> debug.metrics().core.BlockInsertions
{
  Avg01Min: '10 (0.17/s)',
  Avg05Min: '61 (0.20/s)',
  Avg15Min: '168 (0.19/s)',
  Maximum: '2.157261657s',
  Minimum: '2.271716ms',
  Percentiles: {
    20: '6.993756ms',
    50: '12.342836ms',
    80: '21.765944ms',
    95: '218.500479ms',
    99: '376.015984ms'
  },
  Total: '432 (0.22/s)'
}

Or in a raw format for visualization, reporting, processing, monitoring, etc. via debug.metrics(true):

> debug.metrics(true).p2p.InboundTraffic
{
  AvgRate01Min: 1599.6190029292586,
  AvgRate05Min: 5367.754506658111,
  AvgRate15Min: 3761.057607521597,
  MeanRate: 2907.3919382272857,
  Total: 5901154
}
> debug.metrics(true).core.BlockInsertions
{
  AvgRate01Min: 0.18424704676246578,
  AvgRate05Min: 0.19801030732706026,
  AvgRate15Min: 0.18758183634199713,
  MeanRate: 0.22205942695239234,
  Percentiles: {
    0: 2271716,
    1: 2595259.96,
    // ...
    98: 267196822.1600002,
    99: 362098004.77999866,
    100: 2157261657
  },
  Total: 453
}

The PR furthermore introduces a new command to geth, geth monitor, that allows the real time visualization of any of these metrics:

The monitor will plot all requested metrics on individual line charts, expanding user patterns:

• geth monitor core/BlockInsertions/Percentiles/80: a single metric to chart
• geth monitor core/BlockInsertions/Percentiles/25 core/BlockInsertions/Percentiles/95: two metrics, side by side
• geth monitor core/BlockInsertions/Percentiles/25,80,95: three metrics, with a common root
• geth monitor p2p: expand and display all metrics under the p2p collection
• geth monitor eth/db/block,state/Reads,Writes: various metrics with various wildcards (image above)

The charts are automatically scaled below 1000, and the scaling units inserted into the chart labels. The colors of the charts are also updated depending on the scale, ranging in [blue, cyan, green, yellow, red], with blue being none scaled, and every consecutive one a 1000 multiplier.