gdss-brick >> gh17 Redesign disk storage:

Summary:
- Refactoring around squidflash primer. Squidflash primer is now
  triggered by brick_server layer instead of brick_ets layer. This
  will make easier to track request handling path in brick_server.
- do_mod 'insert_existing_value' (used by rename/6 operation) in
  downstream brick now triggers squidflash primer if hlog file for
  OldKey is frozen. However this will be removed by next commit as it
  does not prevent the order of modification events and causes a fatal
  error.

Details:
- brick_server:
  * Now brick_server triggers squidflash primer in brick_ets. Update
    handle_call_via_impl/3 and handle_cast({ch_log_replay_v2, ...})
    to do so.
  * Refactor and simplify impl_details/1.
  * Remove or comment out some debug prints.

- brick_ets:
  * Export some internal functions as bcbs (brick callbacks):
    ** Rename handle_call({do, ...}) to bcb_handle_do/3
    ** Rename squidflash_primer/4 to bcb_squidflash_primer/3
    ** Rename keys_for_priming/3
       to  bcb_keys_for_squidflash_priming/2
       and bcb_keys_for_squidflash_priming/3
  * Refactoring around squidflash_primer:
    ** do_do1b/6 no longer triggers squidflash_primer. This will be
       done by brick_server functions. DoOp and From params in do_do/5
       and do_do1b/6 are no longer needed and removed.
    ** bcb_squidflash_primer/3 and supporting functions no longer
       build resubmitting request by itself. Instead it takes a
       function as second parameter to do that.
  * Remove some old debug prints.
  * Comment out old debug functions and their exports:
    ** sync_pid_start/1
    ** sort_test0/0
    ** slurp_log_chunks/1
    ** debug_scan/1
    ** debug_scan/2
    ** debug_scan2/1
    ** debug_scan2/2
    ** debug_scan3/2
    ** debug_scan3/3
    ** debug_scan4/2
    ** debug_scan4/3