Cleanup SYSCTL RAW OIDs on cleanup

Better cleanup on RAW initialization failure

Thanks for the review, I will fix as stated in the replies.

Rebased on changes.

Simplified raw_buffer free routines to be easily callable from cdev dtors.

In D47429#1090265, @andrew wrote:

What will happen if we read a partial message then close the file? Or if we duplicate the file descriptor?

Removed 2 fields from scmi_softc that needed to be added before in this series.

Added scmi_raw dep and a few missing fields in scmi_softc, mistakenly added in
a following patch

Added missing parenthesis on retirn stm.

Reworked handling of devicetree props

Reworked handling of max_payld-sz and other transport props

Fix mishandled retvalue on scmi_shmem_msg_poll and related funcs

In D47421#1081786, @andrew wrote:

Does scmi_buf_get need to be updated with SCMI_MAX_MSG_REPLY_SIZE changing?

Removed options that enabled SCMI RAW by default (on the wrong file)

Fixed if (token)

Reworked to rebase cleanly on other series changes

Moved sc->trs allocation earlier

Made scmi_detach static

Removed the extres header references

Using mallocarray and dropping empty interface method

Removed needless checks on malloc, using mallocarray

Removed needless malloc checks

Removed check on malloc return value

Removed needless empty interface methods

Removed __FBSDID

Removed the __FBSDID

Thanks for the review @bryanv.
Regarding:

In D43045#989478, @andrew wrote:

Could you use unr(9), the kernel unit number allocator, to allocate the tokens?

In D43044#989468, @andrew wrote:

The attach and detach functions that just call smci_attach and smci_detach could be removed. If a driver doesn't implement a method, but the parent does then the parent implementation will be used.

SCMI: Add support for multiple transports and add VirtIO/SMC transports