fix: HasEntrypoints issues #1385 and #1260

[miparnisari]

Description

Alternative PR to #1405.

This one only fixes the issues with no refactors.

References

• Close no entrypoints defined error when saving a valid model #1385
• Close OpenFGA reports that model is invalid - no entrypoints defined #1260

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.65%. Comparing base (d3a0016) to head (1875f5f).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1422      +/-   ##
==========================================
+ Coverage   85.62%   85.65%   +0.04%     
==========================================
  Files          85       85              
  Lines        7910     7913       +3     
==========================================
+ Hits         6772     6777       +5     
+ Misses        800      799       -1     
+ Partials      338      337       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jon-whit]

Nice changes! Looks solid 👍

We'll want to revisit the case, for example:

type user
type employee

type document
    relations
        define editor: [employee]
        define viewer: [user] and editor  

and

type user
type employee

type document
  relations
    define restricted: [employee]
    define viewer: [user] but not restricted

The question is whether or not these should have entrypoints? In terms of semantics, should a relation involving intersection or exclusion have non-zero entrypoints if the two operands do not share at least one overlapping terminal subject type? My intuition tells me that the models above should have 0 entrypoints.