Linux: O_TMPFILE and insert inode hash rework

[snajpa]

Motivation and Context

Byproduct of investigation done for #16608 - these are the more optional bits, cosmetics, in hope to simplify reasoning about inode lifetime

Description

This change:

• reworks inode hash insert so that its in topmost layer and with inode unlock in the same order as other Linux FS implementations do it

• reworks O_TMPFILE creation to use new linux specific zp->z_is_tmpfile, this fixes a potential unchecked error return value of zpl_xattr_security_init

• changes igrabs to zhold where it can

• exchanges truncate_setsize for truncate_inode_pages_final in zpl_evict_inode to follow vfs.rst's
  "the method has to use truncate_inode_pages_final()"

notes + questions:

• not sure about the change from "working with inode" to "working with znode" in zpl_link, can remove that, that's a bit of a leftover

• I'd like to eliminate the txg_wait_synced in zfs_link in the tmpfile case, I'd log it as a new file creation, is that ok? edit: this probably won't work, or will be too much work, I have another idea:

log the tmpfile creation at creation, log all the changes, acl/xattr/writes, if it comes to replay then put it to unlinked set replay all the changes and then the unlinked set get processed before mount anyway, right?

• I don't feel this is solid enough to be included in release branches yet, it needs vetting from other people with their workloads IMO; can also split this in parts if it makes more sense

How Has This Been Tested?

Localy during development and via the bots also

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Breaking change (fix or feature that would cause existing functionality to change)
• Library ABI change (libzfs, libzfs_core, libnvpair, libuutil and libzfsbootenv)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the OpenZFS code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• I have run the ZFS Test Suite with this change applied.
• All commit messages are properly formatted and contain Signed-off-by.

[behlendorf]

To make this easier to review can you break each of these 4 logical changes in to it's own commit and open a PR for each. Conceptually, these seems to make sense but I'd like to work through them one at a time.

[behlendorf]

Let's leave this as igrab(). It's in the Linux specific code already, and switching it means it does a pointless ITOZ(ZTOI()). Plus the comment above references igrab().

[snajpa]

I'd rather modify the comment if it works for you, then a grep for igrab on the codebase should return the zhold macro and zfs_zget; this use here is exactly what zhold translates to, so I think it would make sense to convert it - and only keep that one igrab in zget as is, because there we actually need the result of the condition.