I've used some of Pat David's film emulations at times past. Downloaded the HaldCLUT.png files directly from the G'MIC webpage:
https://gmic.eu/film_emulation/negative_new_sample_1.shtml#browseJust a matter of clicking on the selected thumbnail image to get the download for the respective HaldCLUT.png file.
Like Dmitry, I used IWLTBAP_LUT_Generator for converting to CUBE LUT's
https://sellfy.com/p/aQ1y/Nothing complicated. Simple one-click conversion.
Just compared some of 'so-converted' LUT's that I kept on file with the equivalent LUT's from Fernando's download pack and they produced identical scope profiles (with some test HD-AVC clips) in Resolve 15.
Edit:
Did that: downloaded the Raw Therapy pack and converted some of the HaldCLUT files to CUBE LUT's in like manner with the IWLTBAP-LUT generator. Testing in Resolve and comparing the profiles obtained with the equivalent LUT's from Fernando's pack (and mine) it is clear that the original have been 'modified' somewhat - generally higher contrast.
Also, as you noted Dmitry, the Raw Therapy HaldCLUT png's are higher resolution (1728 x 1728) than the 'original' G'MIC files (512 x 512) and the IWLTBAP-converted CUBE LUT files are way larger (79.7 MB vs 7.0 MB).
Beyond that, I can't really comment having never used Raw Therapy and knowing next to nothing about the complexities of 3D LUT's.
Edit: According to the ReadMe.txt in the RawTherapee (not Raw Therapy ) pack, the HaldCLUT.pngs are Image Level 12, sRGB, 8bits/channel. As I learn, the 'Image Level' is the cube root of the resolution and equals the number of columns. Which makes the 'original' G'MIC (512) images, Level 8:
http://www.quelsolaar.com/technology/clut.html
http://rawpedia.rawtherapee.com/Film_Simulation
In theory, the higher the Image Level, the greater the precision, but if the RawTherapee HaldCLUT's were (as assumed) derived from the originals, with some modifications along the way, I can't see that would make the transforms any more precise.