Re-enable Big pipe multiple replacements test

Problem/Motivation

In #3390178: big_pipe sometimes fails to load blocks we added a test to big pipe that generates a page with 3000 placeholders. This was the only way we could find to reproduce a race condition seen in the wild.

When we switched to w3c testing 3,000 placeholders caused a timeout (and still does), so we reduced the number to 2000.

2000 is still enough to make the test run very slow - as far as I can tell it is due to chrome choking on the placeholder rendering/replacement when it's CPU constrained although that is mostly conjecture. See #3590076: Speed up BigPipeRegressionTest for some experimentation.

In that issue, we skipped the test because it's more or less impossible for it to fail if we re-introduced the original bug. This issue is for trying to replace/improve it with something that consistently reproduce the bug. Potentially we could also give up and remove the coverage too.

Steps to reproduce

Proposed resolution

Remaining tasks

User interface changes

Introduced terminology

API changes

Data model changes

Release notes snippet