workaround for device lost error on windows + nvidia graphics, use 4k offset alignment for all buffer allocation

[nihui]

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 95.70%. Comparing base (7b63f5d) to head (4718aef).

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #6102   +/-   ##
=======================================
  Coverage   95.70%   95.70%           
=======================================
  Files         827      827           
  Lines      270161   270161           
=======================================
+ Hits       258560   258562    +2     
+ Misses      11601    11599    -2     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

:rocket: New features to boost your workflow:

• :snowflake: Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions[bot]]

The binary size change of libncnn.so (bytes)

architecture	base size	pr size	difference
x86_64	16503040	16503040	0 :kissing_heart:
armhf	7373828	7373828	0 :kissing_heart:
aarch64	10775528	10775528	0 :kissing_heart:

[nihui]

close for https://github.com/Tencent/ncnn/pull/6296