[SYCL][TEST] Remove check_has.cpp dependency on metadata order

[jinge90]

Previously, we committed #9642 aiming to break dependency on metadata order for check_has.cpp but the commit was broken unexpectedly. This PR re-lands it.

[jinge90]

Hi, @steffenlarsen and @maarquitos14
Previously, we added declarations for "hypotf" when oneAPI compiler is used on Windows platform but the change lead to device_has.cpp lit test failure due to meta data order is changed (aspects and srcloc), so we need to fix the test.
Previous test fix is: #9642

[maarquitos14]

Hi, @steffenlarsen and @maarquitos14 Previously, we added declarations for "hypotf" when oneAPI compiler is used on Windows platform but the change lead to device_has.cpp lit test failure due to meta data order is changed (aspects and srcloc), so we need to fix the test. Previous test fix is: #9642

@jinge90 why was it reverted earlier? Did it cause any issue?

[jinge90]

Hi, @steffenlarsen and @maarquitos14 Previously, we added declarations for "hypotf" when oneAPI compiler is used on Windows platform but the change lead to device_has.cpp lit test failure due to meta data order is changed (aspects and srcloc), so we need to fix the test. Previous test fix is: #9642

@jinge90 why was it reverted earlier? Did it cause any issue?

Hi, @maarquitos14
It was not reverted but #10317 changed the test and removed the 2-time checks, our previous test didn't cause any issue. Maybe the author thought that 2-time checks seemed to be unnecessary.
Thanks very much.

[maarquitos14]

Hi, @steffenlarsen and @maarquitos14 Previously, we added declarations for "hypotf" when oneAPI compiler is used on Windows platform but the change lead to device_has.cpp lit test failure due to meta data order is changed (aspects and srcloc), so we need to fix the test. Previous test fix is: #9642

@jinge90 why was it reverted earlier? Did it cause any issue?

Hi, @maarquitos14 It was not reverted but #10317 changed the test and removed the 2-time checks, our previous test didn't cause any issue. Maybe the author thought that 2-time checks seemed to be unnecessary. Thanks very much.

I see, thanks for explaining.

[maarquitos14]

LGTM.

[jinge90]

Hi, @intel/llvm-gatekeepers
Could you help merge this patch?
Thanks very much.