Charlatan speed

First attempts to quantify speed of
charlatan. Instantiating multiple R6 classes is
slow, very visible in fraudster() which initiates
all of the providers.

Relates to #113

tests/testthat/test_benchmarkspeed.R

@@ -0,0 +1,57 @@
+library(microbenchmark)
+mb_summary <- function(x) {
+    res <- summary(x, unit="us")
+    data.frame(name = res$expr, median = res$median)
+}
+benchmarktime = c(1,10,100,1000,10000)
+results<- microbenchmark(
+                    # median
+    ch_job(n=1),    # 761.1   = 761.1
+    ch_job(n=10),   # 998.2   = 99.82
+    ch_job(n=100),  # 2748.9  = 27.49
+    ch_job(n=1000), # 20558.3 = 20.56
+    times = 100L
+)
+jp<-JobProvider_en_US$new()
+
+result2=microbenchmark(           # median
+    replicate(1, jp$render()),    # 58.08    = 58.08
+    replicate(10, jp$render()),   # 222.11   = 22.21
+    replicate(100, jp$render()),  # 1910.45  = 19.10
+    replicate(1000, jp$render()), # 19050.15 = 19.05
+    times = 100L
+)
+
+result3=microbenchmark(           # median
+    fraudster(locale = "en_US"),  # 40771.0
+    JobProvider_en_US$new(),      #   493.4
+    times = 100L
+)
+
+result4 = microbenchmark(
+    ch_norm(n = 1),   # 288.806
+    rnorm(1),         #   4.544
+    ch_norm(n = 100), # 270.063
+    rnorm(100),       #   7.156 (optimized functions)
+    times = 100L
+)
+
+# Yes we pay a performance penalty for using R6. 
+#
+# other notes https://r6.r-lib.org/articles/Performance.html
+# instantiation speed 
+# R6$new()  53.05
+# R6 with private fields 75.85
+# 
+# Instantiating the classes takes a median time of 75.85 ms
+# Instantiating JobProvider_en_US$new(),   takes 493.4 ms
+# 
+# but instantiating fraudster(locale = "en_US"), takes 40771.0 almost 83 times longer.
+# That is a massive class with all sorts of functions. 
+# 
+# creating multiple of a thing is always faster. the relative speedup is especially
+# noticable after 100 repititions.
+# 
+# This package will never be as fast as the optimized functions like rnorm. (while using it)
+# 
+# maybe add a test for all ch_ functions with 100 reps to be within 30 microseconds?