Database table sizes

As of current date, after running the Pioreactor for a few months in a small cluster, here's a snapshot of how large the Sqlite3 artifacts are on disk (in mb) (some of these tables aren't around anymore):

             od_reading_statistics   0.004096
    sqlite_autoindex_experiments_1   0.004096
        led_automation_settings_ix   0.004096
     dosing_automation_settings_ix   0.004096
           led_automation_settings   0.004096
                     sqlite_master   0.012288
        dosing_automation_settings   0.020480
                    stirring_rates   0.020480
temperature_automation_settings_ix   0.024576
                    experiments_ix   0.024576
                       experiments   0.028672
             alt_media_fraction_ix   0.073728
   temperature_automation_settings   0.086016
                      config_files   0.122880
                alt_media_fraction   0.180224
                  dosing_events_ix   0.217088
                       pid_logs_ix   0.335872
                     dosing_events   0.692224
                          pid_logs   1.441792
           temperature_readings_ix   1.949696
              temperature_readings   4.399104
                           logs_ix  21.004288
           od_readings_filtered_ix  22.425600
              od_readings_filtered  45.903872
          kalman_filter_outputs_ix  50.630656
                   growth_rates_ix  50.630656
                              logs  91.009024
                      growth_rates 105.705472
                od_readings_raw_ix 158.986240
                     led_events_ix 164.225024
                   od_readings_raw 367.345664
                        led_events 450.797568
             kalman_filter_outputs 638.550016

How can we optimize this? What's redundant?

• led_events data is very redundant - this was only holding the IR LED blinking on and off over experiments.
• Probably remove the index on led_events - we don't use it in the UI
• Remove the index kalman_filter_outputs_ix
• kalman_filter_outputs redundancy is storing arrays of floats as strings. That's silly. We could truncate the precision of the floats:

[0.9817443680619926, 1.007342729613004, 4.877678040174688e-05, 0.0]

to

[0.9817, 1.00734, 4.87e-05, 0.0]

No files or tests are named. Start by locating the SQLite schema and the tables listed in the issue, then measure current table and index sizes; done means the remaining redundant storage is reduced and the database behavior is validated without breaking existing use.