[cache] Providing a mechanism for disabling the datasource/database cache

[john-bodley]

This PR updates the cache-timeout delegation logic which previously checked whether the datasource or database cache_timeout evaluated to True rather than whether it was None. The distinction here is you should be able to set a cache-timeout of zero at the datasource/database level to indicate that the slice will not timeout.

I've also updated the form descriptions which explicitly mention the timeout is in seconds and what the fall back logic is, i.e.,

chart:

table:

database:

datasource:

cluster:

to: @graceguo-supercat @michellethomas @mistercrunch @timifasubaa

[codecov-io]

Codecov Report

Merging #5315 into master will increase coverage by <.01%.
The diff coverage is 50%.

@@            Coverage Diff             @@
##           master    #5315      +/-   ##
==========================================
+ Coverage   61.33%   61.33%   +<.01%     
==========================================
  Files         369      369              
  Lines       23483    23484       +1     
  Branches     2713     2713              
==========================================
+ Hits        14404    14405       +1     
  Misses       9067     9067              
  Partials       12       12

Impacted Files	Coverage Δ
superset/views/core.py	72.96% <ø> (ø)	⬆️
superset/connectors/sqla/views.py	70.47% <ø> (ø)	⬆️
superset/viz.py	81.35% <0%> (ø)	⬆️
superset/connectors/druid/views.py	68.24% <100%> (+0.21%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 16d2633...662874b. Read the comment docs.

[mistercrunch]

I think it's possible for cache_timeout to be an empty string too (maybe once it's set and cleared on the CRUD), I think that needs to be explicitly handled and tested too.

Also another good thing to do would be adding info in the help text which I think has been missing. We should make it clear we're expecting seconds and that 0 turns off the cache.

[john-bodley]

@mistercrunch cache_timeout is an int and thus is NULL in the database if cleared in the CRUD view. The one caveat is -1 will evaluate to True.

[john-bodley]

I can update the form description. Note 0 doesn't disable the cache but really ensures that the cache will always miss.

[mistercrunch]

LGMT

[john-bodley]

@mistercrunch I've updated the description in the FAB forms which should provide more transparency around caching and the fall back logic.

[mistercrunch]

LGTM