Intern BinaryPacker.Element and EntityData values, and PlaybackData animation ids.

[JaThePlayer]

Saves ~50MB RAM and reduces allocated object count by ~2 millions with Strawberry Jam.

More results starting from here: https://discord.com/channels/403698615446536203/429775439423209472/1533091706080067657

[DashingCat]

As discussed on Discord, this PR initially broke the "Fast Level Loading" (FLL) PR #1072, but this was taken into account in the latest commits.

Besides the RAM save, this PR has an impact on loading performance:

Without changes, FLL disabled (current state):
 - LEVELS LOAD: 8829ms
 - LEVELS LOAD: 8696ms
 - LEVELS LOAD: 8646ms

With changes, FLL disabled:
 - LEVELS LOAD: 8799ms
 - LEVELS LOAD: 8956ms
 - LEVELS LOAD: 9002ms

Without changes, FLL enabled:
 - LEVELS LOAD: 7072ms
 - LEVELS LOAD: 7172ms
 - LEVELS LOAD: 7186ms

With changes, FLL enabled:
 - LEVELS LOAD: 8452ms
 - LEVELS LOAD: 8247ms
 - LEVELS LOAD: 8169ms

While this PR introduces a slowdown while loading levels, assuming we can merge both this PR and #1037, we should still have better both loading times and RAM usage, so I approve this PR.

[maddie480-bot]

The pull request was approved and entered the 3-day last-call window.
If no further reviews happen, it will end on Aug 17, 2026, 7:50 PM UTC, after which the pull request will be able to be merged.

[wuke32767]

here is my test result (only strawberry jam):

Slow Level Loading
 - LEVELS LOAD: 3046ms
 - LEVELS LOAD: 3165ms
 - LEVELS LOAD: 3074ms
Fast Level Loading
 - LEVELS LOAD: 5281ms
 - LEVELS LOAD: 4267ms
 - LEVELS LOAD: 4116ms
 - LEVELS LOAD: 4513ms
 - LEVELS LOAD: 4211ms

i'm trying two ways to optimize it a little, by replace the ConcurrentDictionary with _, but failed:

Fast Level Loading (Immutable)
 - LEVELS LOAD: 4556ms
 - LEVELS LOAD: 4536ms
 - LEVELS LOAD: 5490ms
 - LEVELS LOAD: 4158ms
 - LEVELS LOAD: 4539ms
Fast Level Loading (ThreadStatic)
 - LEVELS LOAD: 4272ms
 - LEVELS LOAD: 4608ms
 - LEVELS LOAD: 4408ms

[wuke32767]

wait, i think i should nitpick it.
the FLL compatible stuff should be removed for now and opened in a new pr (or in FLL pr). currently this pr makes FLL slower.
[non-blocking]

[maddie480-bot]

The last-call window for this pull request ended. It can now be merged if no blockers were brought up.