Jonas Wanke

In the LIR optimization, we should drop values as early as possible and not at the end to reduce memory usage

Hopefully, this prevents a few use-after-frees in the future

### Contact Details _No response_ ### Input ```candy newline := " " ``` ### Actual Output ```candy newline := " " ``` ### Desired Output ```candy newline := " "...

### Contact Details _No response_ ### Input ```candy Foo % Foo -> [ Key: Value, ] ``` ### Actual Output ```candy Foo % Foo -> [ Key : Value ,...

```[tasklist] ### Common - [ ] HTTP: IP Address, Port, Socket modules ``` ```[tasklist] ### Server - [ ] https://github.com/candy-lang/candy/issues/788 ``` ```[tasklist] ### Client - [ ] HTTP: Basic Client...

```[tasklist] ### Tasks - [ ] #619 - [ ] #657 - [ ] #775 - [ ] #658 - [ ] #654 - [ ] #691 - [ ]...

Currently, we hard-code paths for the linker (targeting x86-64 Linux). These differ by OS and CPU architecture