Excessive memory usage caused by unique synthetic NFGs in string

[Maxdamantus]

Using unique graphemes containing 2 or more code points seems to result in memory complexity that is quadratic to the number of graphemes. Example with 50,000 unique graphemes:

$ raku -e 'for 0..50000 { my @a = $_, 0x300; print @a.chrs; };' >test.txt
$ ls -lh test.txt 
-rw-r--r-- 1 max max 242K Jan 17 22:34 test.txt
$ raku -e 'slurp;' <test.txt
Killed
$ raku -v
Welcome to 𝐑𝐚𝐤𝐮𝐝𝐨™ v2021.09.
Implementing the 𝐑𝐚𝐤𝐮™ programming language v6.d.
Built on MoarVM version 2021.09.

# dmesg | grep 'Out of memory' | tail -1
[106097.011005] Out of memory: Killed process 16841 (raku) total-vm:15931556kB, anon-rss:15799148kB, file-rss:4kB, shmem-rss:0kB, UID:1000 pgtables:31092kB oom_score_adj:0

Memory consumption seems to be approximately 8.5*(graphemes/1000)^2 MiB, so 44k graphemes will exhaust the memory on my 16 GiB system.

[MasterDuke17]

Here's where heaptrack says the memory is being allocated (for 30k so it wouldn't get killed).

Welcome to Rakudo™ v2021.12-118-g888f05d25.
Implementing the Raku® Programming Language v6.d.
Built on MoarVM version 2021.12-67-g4b2de6d1d.