Tile QA simplified

[julienguy]

Implemented a "simplification" of the tile QA.
(after discussion with Stephen and David)

 - exposure_goodfiber = (QAFIBERSTATUS&fibermask.mask(bad_qafstatus_mask)==0) , per fiber and exposure
 - tile_fiber_efftime = sum(efftime[exposure_goodfiber]) , per fiber
 - tile_goodfiber = ( tile_fiber_efftime > 0.9*mintfrac*goaltime ) ,  (param fiber_rel_mintfrac = 0.9) , per fiber
 - tile_always_goodfiber = OR(exposure_goodfiber) , per fiber 
 - tile_efftime = median(tile_fiber_efftime[tile_always_goodfiber]) , per tile
 - tile_is_valid = ( tile_efftime > mintfrac*mintfrac*goaltime ) & ( sum(tile_goodfiber) > 3000 ) , (param  min_number_of_good_fibers = 3000)

Automatic validation ('VALID' in header of tile_qa.....fits), can be overridden by visual inspector (that always
has to validate). Overridden entries are noted (if someone wants to double check).

[coveralls]

Coverage decreased (-0.07%) to 26.445% when pulling da35617 on tile_qa_simplified into 8a4e68a on master.

[julienguy]

Example

desi_tile_vi --prod daily --qa . --night 20210514:20210618 -o mytiles-new.csv -u jguy --viewer display
...
Showing tiles/cumulative/20240/20210514/tile-qa-20240-20210514.png ...
The code thinks it's a valid tile
PROMPT: Is tile 20240 a valid tile? (y/n/save/exit/pass/help)

>>> n
USER ENTERED >>> n
Your have overridden the automated validation.

[julienguy]

(I actually agree it is a valid tile :-)

[sbailey]

I did several small updates, but overall looks good. I generated tile QA for all main survey tiles in /global/cfs/cdirs/desi/users/sjbailey/test/tiles/cumulative . Let's vet those a bit more and then we can merge and rerun for the real daily prod (it takes ~10 min with --nproc 32 on an interactive node).

One desi_tile_vi hiccup: the input csv file has default values for QA and USER and "none" so they get read in as a 4-character table, which works fine for "jguy" but longer names like "stephen" get truncated to "step".

In desihub/desitarget#755 @geordie666 expects the new input to be tiles-specstatus.ecsv which gives the opportunity to define the column types of the ascii table to allow more characters, but requires a bit of bootstrapping to get it right. A topic for tomorrow...