apply functions over windows into weighted data

`rolling`, `tiling`, and `running` all provide data weighting.

The functions for weighted data follow the unweighted function signatures.
- the weighting is given after the data, as the last positional arg

Weighting for a data vector is given as one of the subtypes of StatsBase.AbstractWeights

To use myweights::Vector{<:Real} as weights

When a window is provided weights, is important to normalize those weights. For most applications, the weights should sum to 1.0 (or just less than 1.0).

The function safeweights(weights::AbstractVector{<:AbstractFloat}) does the this with care. The sum of a weight vector returned from safeweights is assured to be <= 1.0 while staying within a small multiple ofeps(eltype(weights)).

• scale the values so they sum to 1.0 (or a few eps less than 1.0)
  • myweights1 = safeweights(myweights1)
• convert the values to StatsBase.AnalyticWeights (usually)
  • weighting = AnalyticWeights(myweights1)

using weights

See also: safeweights, datastreams