rhel8/9: data corruption due to stale cached iomaps

Issue

We found that some of files on xfs were partially filled by zero unexpectedly. We created files by filling them from multiple threads without overwrite/overlap on each writes using pwrite(2):

    thread1  thread2  thread3
       |     |        |
       |     |        +-----+--------------------------+----------------- ...
       |     |              |                          |
       |     +-----+--------------------------+--------------------------+---- ...
       |           |        |                 |        |                 |
       +-----+-----------------------+--------------------------+----------- ...
             |     |        |        |        |        |        |        |
     write   v     v        v        v        v        v        v        v
    file [   |     |        |        |        |        |        |        |        | ...

    (written by:
             11111122 22222233 33333311 11111122 22222233 33333311 11111122 222222 ...)

     With this write pattern, corruption was observed, for example:
             11111122 22222233 33333311 11111122 00000033 33333311 11111122 222222 ...)
                                                 ^^^^^^