Draft: PersistentDataModel: Using to_chars instead of sprintf

This version improves performance of these methods between 1.8x to 3x.

This takes some nice elegant code and makes it very ugly however. So I will leave it to @gemmeren to decide if they want to use it.

cc @smh

Proof of concept can be seen in this test code: https://cernbox.cern.ch/s/J8k346RKzyrTdch

Run on (24 X 3700.04 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x12)
  L1 Instruction 32 KiB (x12)
  L2 Unified 512 KiB (x12)
  L3 Unified 32768 KiB (x1)
Load Average: 0.05, 0.05, 0.01
--------------------------------------------------------------
Benchmark                    Time             CPU   Iterations
--------------------------------------------------------------
BM_TokentoStringNEW       92.9 ns         92.9 ns      7553115
BM_TokentoStringOLD        205 ns          205 ns      3402696
BM_keyNew                 71.0 ns         71.0 ns      9817658
BM_keyOLD                  118 ns          118 ns      5911308
toStringNEW               62.3 ns         62.3 ns     11244709
toStringOLD                102 ns          102 ns      6849851