hisilicon-osdrv-hi3519v101: rmmod openhisilicon source modules by open_* name

[widgetii]

Summary

load_hisilicon for hi3519v101 issues rmmod -w against vendor file names (hi_sensor_i2c, hi_sensor_spi, hi_osal, hi_pwm, hi_piris, hi_mipi) for modules openhisilicon ships with KBUILD_MODNAME=open_*. Those rmmods miss silently with "No such file or directory", and the failure cascades — hi_sensor_i2c (loaded as open_sensor_i2c) holds a reference into hi3519v101_isp, so rmmod -w hi3519v101_base later fails with "Resource temporarily unavailable" because base is still pinned by isp.

Symptom in CI

The QEMU CI on OpenIPC/openhisilicon#63 for hi3519v101 shows this directly:

```
rmmod: can't unload module 'hi_sensor_i2c': No such file or directory
rmmod: can't unload module 'hi3519v101_base': Resource temporarily unavailable
rmmod: can't unload module 'hi_osal': No such file or directory
```

Verification

Extracted .gnu.linkonce.this_module from the shipped Apr 28 latest 3519v101 tarball:

```
hi_osal.ko → open_osal
hi_sensor_i2c.ko → open_sensor_i2c
hi_sensor_spi.ko → open_sensor_spi
hi_pwm.ko → open_pwm
hi_piris.ko → open_piris
hi_mipi.ko → open_mipi_rx
sys_config.ko → open_sys_config
```

…vs Strategy B blob-wrapped modules that retain vendor names (hi3519v101_base.ko → hi3519v101_base, hi3519v101_sys.ko → hi3519v101_sys, hi3519v101_isp.ko → hi3519v101_isp, …), which is why those rmmod -w calls don't change.

Scope

8 rmmod call sites updated. Insmod calls operate on file names and don't need to change.

Refs

• OpenIPC/openhisilicon#62 — root-cause investigation.
• OpenIPC/openhisilicon#63 — CI assertions; the hi3519v101 job is currently RED and will turn GREEN once openipc/firmware re-cuts `releases/latest` with this script.
• Sibling cv500 fix: OpenIPC/firmware#2026.
• Sibling cv100 fix: opened in parallel.

Test plan

• CI on this PR passes the 3519v101 build path.
• After merge + new `releases/latest` tarball: re-run OpenIPC/openhisilicon#63 and confirm QEMU boot (hi3519v101) flips RED → GREEN.
• On real hi3519v101 / hi3516av200 hardware (separate validation, not in lab here): confirm load_hisilicon -i -sensor <real_sensor> cleanly loads, /dev/sys exists, and majestic serves a frame.