Optimize impcnvtab for newCTFE

[UplinkCoder]

Within the inner functions the 23232 byte large impcnvtab is used,
because of how delegates and closures are implemented in newCTFE
the full copy has to be done everytime the table is touched.
Therefore I opted to pass a pointer explicitly,
which improves compiletimes for newCTFE by a factor of 500

[dlang-bot]

Thanks for your pull request, @UplinkCoder!

Bugzilla references

Your PR doesn't reference any Bugzilla issue.

If your PR contains non-trivial changes, please reference a Bugzilla issue or create a manual changelog.

Testing this PR locally

If you don't have a local development environment setup, you can use Digger to test this PR:

dub fetch digger
dub run digger -- build "master + dmd#9112"

[thewilsonator]

Nice! What are the numbers for old CTFE?

[UplinkCoder]

Haven't checked, probably little to nothing.

I don't know how they implement delegates.

[kinke]

This should most likely be the other way around - optimizing newCTFE instead, to prevent this ugliness and slowness (factor of 500?!) for similar code.

[UplinkCoder]

currently I cannot devote resources to optimizing border cases like giant structs being passed via closure variables. It's quite hard to fix actually since newCTFE has no notion of a stack.

[kinke]

I fully understand, but I don't see a reason for this to be in upstream DMD, it just uglifies the code unless someone can show it improves the current interpreter performance too.

[thewilsonator]

currently I cannot devote resources to optimizing border cases like giant structs being passed via closure variables. It's quite hard to fix actually since newCTFE has no notion of a stack.

border cases? Hello regex?

[Geod24]

This PR doesn't improve the code readability, and I don't think the change carries its weight.