Decoding speedup

[baltzell]

• Zero change to resulting output data
• More caching, more hashing, less looping
• Replace string comparisons with DetectorType comparisons
• Replace string IndexedTable access with column indices
• Remove safety checks on hash-based accessors
  and
• Moveinstall-clara to bin
• Add Clas12Reader CLARA service

See #949 for finer commits (had trouble rebasing across submodule changes).

Decoding for RG-D LD2 is now down to ~3 ms per event on ifarm23, my laptop, and GitLab CI jobs at JLab.

[baltzell]

hipo-diff reports success on clas_018779.evio.00001 for all the decoded banks below. But the filesize isn't identical, hmm (ifarm2402:/scratch/baltzell/*.hipo).

|                      BAND::adc |                                                |    644 (kb)
|                      BAND::tdc |                                                |    702 (kb)
|                       BMT::adc | ░░░░░░░░░░░░░░░░░░░░░░░░░░                     | 419687 (kb)
|                       BST::adc | ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░           | 576971 (kb)
|                       CND::adc | ░                                              |  20247 (kb)
|                       CND::tdc |                                                |  10015 (kb)
|                      CTOF::adc |                                                |   8559 (kb)
|                      CTOF::tdc |                                                |  10043 (kb)
|                     DC::jitter | ░░░░░░                                         | 101497 (kb)
|                        DC::tot | ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░ | 723945 (kb)
|                      ECAL::adc | ░░░                                            |  61329 (kb)
|                      ECAL::tdc | ░░                                             |  39519 (kb)
|                       FMT::adc | ░░░░░░                                         |  98624 (kb)
|                      FTOF::adc | ░░░░                                           |  73456 (kb)
|                      FTOF::tdc | ░░░                                            |  54868 (kb)
|                       HEL::adc |                                                |   3865 (kb)
|                   HEL::decoder |                                                |    139 (kb)
|                      HEL::flip |                                                |      4 (kb)
|                    HEL::online |                                                |    745 (kb)
|                    HEL::scaler |                                                |      8 (kb)
|                      HTCC::adc |                                                |   2840 (kb)
|                      LTCC::adc |                                                |   3812 (kb)
|                      LTCC::tdc |                                                |   1093 (kb)
|                       RAW::adc |                                                |   2219 (kb)
|                     RAW::epics |                                                |      4 (kb)
|                    RAW::scaler |                                                |     86 (kb)
|                       RAW::tdc |                                                |   5179 (kb)
|                       RAW::vtp | ░░░░░░░░░░░░░░                                 | 222818 (kb)
|                        RF::adc |                                                |   2833 (kb)
|                        RF::tdc | ░                                              |  16663 (kb)
|                      RICH::tdc | ░░░                                            |  48323 (kb)
|                    RUN::config |                                                |   3436 (kb)
|                    RUN::scaler |                                                |      6 (kb)
|                   RUN::trigger |                                                |   2386 (kb)

[baltzell]

Ah, the expanded total event size (and count) from hipo-utils is identical.

[baltzell]

Oh, oops, RASTER::adc is missing. Checking for a CCDB duplicate ...

[baltzell]

breaking this out into separate pull requests