Solidify a consistent interface for running power flows

[GabrielKS]

Here are the extant claims about how the user can run a power flow using PowerFlows.jl:

• Function solve_powerflow is exported, it can run lots of types of power flows and outputs a possibly nested Dict of DataFrames
• Function solve_ac_powerflow! is exported, it does the same calculation as solve_powerflow(::ACPowerFlow, ...) but updates the System with the results rather than returning them
• The docs mention functions run_powerflow and run_powerflow! but these do not exist

My first thought for how to clean this up is to just have a pair of functions solve_powerflow, solve_powerflow! (or run_powerflow, run_powerflow!) where the non-! version returns results and the ! version updates the System.

Also there is an exported function write_results whose PF methods always take a PowerFlowData, but there does not seem to be a way for one to get a populated PowerFlowData from any exported function. We should think about whether/how we want the user to be interacting with PowerFlowData and adjust the interface accordingly.

Another inconsistency is that solve_powerflow(::DCPowerFlow, ...) gives results indexed by time (currently only "1") and solve_powerflow(::ACPowerFlow, ...) does not.

[GabrielKS]

Though we are not currently planning for end users to interact with the PowerFlowData struct, we need to have a stable function for solving a power flow in-place given a PowerFlowData for use elsewhere in Sienna, like in the PowerSimulations power flow in the loop feature. Right now that function is called solve_powerflow! and it isn't exported.