tests: extend testatomic timeout to 120 seconds

[pinotree]

Description

The default (i.e. 10 seconds) is not enough for slower architectures, such as alpha or hppa.

Raise it to 120 seconds according to the feedback of hppa porters: https://bugs.debian.org/1095774

[madebr]

SDL3 has SDLTEST_TIMEOUT_MULTIPLIER, useful for slower platforms.

[pinotree]

SDL3 has SDLTEST_TIMEOUT_MULTIPLIER, useful for slower platforms.

Yes, I know it exists, and it already used in the Debian packaging, exactly to help all the architectures different than x86/x86_64/arm64.

Since there are already tests that have their own NONINTERACTIVE_TIMEOUT set, this PR here is to support also testatomic OOTB on more architectures (even slower ones) without the need to raise the timeout for all the tests at once.

@madebr Since you requested changes: would you please describe me which actionable steps you'd do with this PR? Thanks in advance!

[madebr]

Since there are already tests that have their own NONINTERACTIVE_TIMEOUT set, this PR here is to support also testatomic OOTB on more architectures (even slower ones) without the need to raise the timeout for all the tests at once.

How slow are we talking? 10s was chosen as a default as testatomic on a 10+ years old computer takes around 1 seconds. Going from 10s to 120s is a big step-up for something that should be fast.

[smcv]

For context, the DEC Alpha and HP PA-RISC (HP/PA) architectures were both considered by their last manufacturer (HP) to be superseded by the Itanium, which was itself superseded by x86_64. According to Wikipedia, new alpha machines were most recently sold in 2007 (with CPUs from 2004 or older), and new hppa machines in 2013 (with CPUs from 2005). These are not recent or fast machines, and I don't think anyone would intentionally choose them for gaming, except perhaps to be able to say that they can.

[smcv]

According to the Debian bug,

Atomic operations are very slow on hppa. The testatomic runtime is about 80 seconds on 4-way 800 MHz rp3440.

[madebr]

According to the Debian bug,

Atomic operations are very slow on hppa. The testatomic runtime is about 80 seconds on 4-way 800 MHz rp3440.

Thank you. I did not realize there was an underlying Debian bug report. Sorry about that.

atomics need to be fast for SDL3 to be performant on today's machines. Increasing the timeout makes the test less useful.

How about setting it to 120 only if HPPA (variable set by cmake/sdlplatform.cmake) is true?