fix: resolve cpu load issue of prepare_uninitialized_buffer

[quake]

After doing some basic profiling, I found that a lot of CPU was spent on memset, and traced the code to find that it was caused by not implementing the prepare_uninitialized_buffer method

below is the bench result comparison

After counting 10 cycles, estimated total time spent on task "10mb_benchmark_with_secio" is 3.0800651s
task name: 10mb_benchmark_with_secio
cycles: 100
total cost: 3.019923503s
average: 30.199235ms
median: 30.015518ms
max: 42.400898ms
min: 28.272243ms

v.s

After counting 10 cycles, estimated total time spent on task "10mb_benchmark_with_secio" is 1.707217s
task name: 10mb_benchmark_with_secio
cycles: 100
total cost: 1.75304821s
average: 17.530482ms
median: 17.24187ms
max: 35.465816ms
min: 16.76556ms

[driftluo]

Yes, I did not notice this implementation. In our implementation, it is guaranteed that we will not read any data in the undefined buffer. This guarantee can be skipped directly here.