Fix problem where we were over eager with disposing variables instances

[nstrayer]

Tried to fix leaks by disposing a variables instance instead of just removing a reference to it. Caused wonky data explorer bug.

QA Notes

Steps to reproduce previously broken behavior

• run Data_Frame <- mtcars
• double click Data_Frame in variables and make sure it opens
• open a new editor
• again double click Data_Frame in variables and make sure it opens in original editor, not a new one.

[github-actions]

E2E Tests 🚀  ^?
This PR will run tests tagged with: @critical

[midleman]

Triggered workflow that runs the culprit test 10x in a row in CI: https://github.com/posit-dev/positron/actions/runs/12304954334. Let's see what happens. 🤞

Update:
Hmm, appears to be very flaky in the same way: https://d38p2avprg8il3.cloudfront.net/playwright-report-12304954334-23403/index.html

[midleman]

@nstrayer, a good way to test if the fix worked locally is to run the e2e test repetitively:
yarn e2e --grep "C701143" --repeat-each 10

I just tried that ^ locally and it passed, however the CI runs behaved differently (as linked above). 🤔