dateReleased field is not respected when marking regressions.

### Environment

SaaS ([https://sentry.io/]())

### Version

*No response*

### Link

*No response*

### DSN

*No response*

### Steps to Reproduce

1. Create release A
2. Create release B
3. Have an issue that has events from both releases
4. Finalise release B
5. Finalise release A
6. Mark the issue as resolved on release A
7. Send a new event to the issue marked as coming from release B

### Expected Result

Altough release B was created after release A, release A was finalised after (populating the `dateReleased`) field.

Thus, events from release B should not trigger a regression.

Per docs:
By default a release is created “unreleased”. Finalizing a release means that we populate a second timestamp on the release record, which is prioritized over date_created when sorting releases in [sentry.io](). Release finalization (and the timestamp) also affects:

* What counts as "the next release" for resolving issues
* What release is used as the base for associating commits if you use --auto

The above is also not respected, as release B still shows as more recent as release A.

### Actual Result

A regression is triggered.


On my tests I used:

* A= 5.0.5.7.0
* B= 5.0.5.6.1

SEMVER is not taking part in this, 5.0.5.6.1 is still shown as more recent. Only Sentry-cli seems to respect the dateReleases (finalised) field: