dimod dismisses part of the expression when both lower bound and higher bound are defined in a single line

**Description**
dimod does not parse constraints correctly when we define both lower bound and higher bound in a single line. It dismisses the first part of the expression.

**Steps To Reproduce**

```
import dimod

v0, v1 = dimod.Integers('xy')

cqm = dimod.ConstrainedQuadraticModel()
cqm.set_objective(v0 + v1)
cqm.add_constraint(5 <= v0 <= 100)

print(cqm.constraints)
```

Output:
`{'c08e0e6': Le(ConstraintView({'x': 1.0}, {}, 0.0, {'x': 'INTEGER'}), 100.0)}`

**Expected Behavior**
The above expression should give the same output as:
```
import dimod

v0, v1 = dimod.Integers('xy')

cqm = dimod.ConstrainedQuadraticModel()
cqm.set_objective(v0 + v1)
cqm.add_constraint(v0 >= 5)
cqm.add_constraint(v0 <= 100)

print(cqm.constraints)
```

Output:
`{'cb82aca': Ge(ConstraintView({'x': 1.0}, {}, 0.0, {'x': 'INTEGER'}), 5.0), 'cdcf213': Le(ConstraintView({'x': 1.0}, {}, 0.0, {'x': 'INTEGER'}), 100.0)}`

**Environment**
- OS: macOS 12.5.1
- Python version: Python 3.10.7

Start by running the reported Python reproduction with dimod.ConstrainedQuadraticModel.add_constraint and inspect how chained comparisons are interpreted. Add a regression test covering 5 <= v0 <= 100 and compare it with two separate constraints. Done means both bounds remain as constraints rather than the lower bound being discarded.

Written by the indexing model from the issue text.