Upstreaming a patch for handling the difference between Linux & Solaris
/proc that we've been carrying in the Solaris packages for 2 years now:

https://github.com/oracle/solaris-userland/blob/master/components/desktop/espeak-ng/patches/03-procfs.patch

This gets rid of the warning message:
```
mbrowrap error: /proc is unaccessible: No such file or directory
```

Signed-off-by: Alan Coopersmith <[email protected]>

Co-authored-by: adr-adr <adr@github>

https://github.com/espeak-ng/espeak-ng/issues/1972

https://github.com/espeak-ng/espeak-ng/issues/1972

Co-authored-by: Baybars Külebi <[email protected]>
Co-authored-by: gullabi <[email protected]>
Co-authored-by: mllopartbsc <[email protected]>

On non-FHS systems like NixOS, there is no `/usr`. Additionally, Nix has
immutable packages so mbrola cannot be installed to espeak-ng’s
`datadir`. Instead, packages are expected to find data under one of the
paths listed in `XDG_DATA_DIRS` environment variable, as defined in the
[XDG Base Directory
Specification](https://specifications.freedesktop.org/basedir-spec/basedir-spec-0.6.html#variables).

Fixes: https://github.com/espeak-ng/espeak-ng/issues/1779

On non-FHS systems like NixOS, there is no `/usr`. Additionally,
Nix has immutable packages so mbrola cannot be installed to espeak-ng’s
`datadir`. Instead, packages are expected to find data under one of the paths
listed in `XDG_DATA_DIRS` environment variable, as defined in the XDG Base
Directory Specification:
https://specifications.freedesktop.org/basedir-spec/basedir-spec-0.6.html#variables

Co-authored-by: Juho Hiltunen <[email protected]>

84e967c0a2 ("Update translate.c") (sic) dropped the initialization of
number_buf, breaking the CI on ubuntu amd64 clang+msan, ubuntu
amd64/i386 gcc/clang+valgrind,

While this probably deserves cleaning up, we'd rather not break the CI,
to be able to properly track regressions.

84e967c0a2 ("Update translate.c") (sic) dropped the initialization of
number_buf, breaking the CI on ubuntu amd64 clang+msan, ubuntu amd64/i386
gcc/clang+valgrind,

While this probably deserves cleaning up, we'd rather not break the CI,
to be able to properly track regressions.

in general american, for and four are homophones, see
https://en.wiktionary.org/wiki/four#Pronunciation_2
https://en.wiktionary.org/wiki/for#Pronunciation

This is due to the horse-hoarse merger, see

https://en.wikipedia.org/wiki/English-language_vowel_changes_before_historic_/r/#Horse%E2%80%93hoarse_merger

This PR accompanies for this merger on the pronunciation level.

Data for the constructed art language Xextan, adapted for current
version of eSpeak-NG.

I added initial support for Faroese language. Did the rules for most
common phonetic rules that are in place for Faroese, and the ability to
more easily make improvements to the phonemes, rules, and exceptions.
Current status is early testing, but voice is mostly understandable at
this point, and pronounces numbers by the correct rules now.

Fixes #890 #616 

essentially reverts this commit
e07002a1c9

Fixes #616 #890

resolve conflict with main

Add more consonant combinations to fix gaps in mb-br1 voice
Look #1054

Adding an initial support for [Karakalpak
language](https://en.wikipedia.org/wiki/Karakalpak_language)

This function was added to the `translate.h` header in #1811 but the
associated function definition was not added to `translate.c`