improve diagnostics in recalib()

[GregorDeCillia]

give the client of recalib() more flexibility about how live diagnostics in ipf() are produced. Add a new callback class surveysd_reporter and two implementations for recalib() and ipf().

ipf(..., reporter = ipf_reporter())
recalib(..., reporter = recalib_reporter())

Those reporters are callbacks that will be invoked after every iteration step of ipf()

# body of ipf()
while (!converged) {
  converged <- ipf_step(data) # data is updated by reference
  reporter(data)
}

Possible parameters

ipf_reporter(freq = 10, show_table = TRUE, show_number_steps = TRUE)
recalib_reporter(freq = Inf, show_table = FALSE, show_number_steps = TRUE)

• freq: how often should the reporter be used? Default: every 10th iteration. Inf menas the reporter is called once after the last iteration.
• show_table: Show a verbose table that characterizes problematic groups?
• show_number_stps: Show a message of the form "not converged after XY steps" or "converged after XY steps"?