feat: Deleted time added to the database

[Monirzadeh]

This PR will record deleted time with bookmark id in a seprate table. any specific unit test needed?

Why we need this?

for create a sync functionality we need to have 3 things created_at modifieded_at and deleted_at the last one added in this PR. as deleted_at use case is just for sync than keep that in separate table make things cleaner.

after merge #894 #896 and #903 i can finally send PR for add sync functionality to Shiori (some part of code is ready in local system). so user can sync their apps with Shiori server.

[codecov[bot]]

Codecov Report

Attention: Patch coverage is 0% with 30 lines in your changes are missing coverage. Please review.

Project coverage is 30.56%. Comparing base (595cb45) to head (aaf0d16). Report is 34 commits behind head on master.

Files	Patch %	Lines
internal/database/mysql.go	0.00%	10 Missing :warning:
internal/database/pg.go	0.00%	10 Missing :warning:
internal/database/sqlite.go	0.00%	10 Missing :warning:

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #903 +/- ## ========================================== + Coverage 25.58% 30.56% +4.97% ========================================== Files 47 60 +13 Lines 5628 4486 -1142 ========================================== - Hits 1440 1371 -69 + Misses 3989 2893 -1096 - Partials 199 222 +23 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

[fmartingr]

I'm still not sold on this because it changes a foundational thing about Shiori and it may be confusing for users (that delete no longer means delete). I'd rather think in a better way for third-parties to know when some bookmarks are gone. Is knowing when a bookmark has been deleted the only remaining problem?

[Monirzadeh]

yes knowing deleted item is the remaining part. i thinking about different method

1. normally programs done thumb stone that i don't like that at all. (terrible method) big NO to me :no_good:
2. a way is that user send id that have in client side and we check one by one (that i think it is intensive unless we find a way to done that quickly)
3. current method is the best way that i found yet that keep minimal possible footprint. not prefect.

client may don't have all bookmarks just have part of it. for example in a database with 1000 bookmark client maybe have just 100 recently item (900 - 1000) but maybe client has bookmark id 503 (get and save that from tag search for example). what is the best method to define which one of that 101 (900-1000 plus bookmark id 503) item deleted?

i need to return something like this:

• bookmarks added after timestamp user send (client last sync time)
• bookmark modified after timestamp user send
• list of id that deleted (better after timestamp user send)

one way is that i return any missing id with something like this (it is a quick code and has some bug) what do you think?

WITH RECURSIVE range(id) AS (
  SELECT 1
  UNION ALL
  SELECT id + 1 FROM range WHERE id < (SELECT MAX(id) FROM bookmark)
)
SELECT id
FROM range
WHERE id NOT IN (SELECT id FROM bookmark);