Reduce memory usage by using rref instead of move-copy

[kghost]

Using rvalue reference can save a call to move constructor.

This PR is trying to replace some arguments of PacketBufferHandle to rvalue reference, in order to check if we can save memory.

[Damian-Nordic]

Nice result! :)

[woody-apple]

@saurabhst @andy31415 ?

[woody-apple]

@kghost Sadly a conflict now, sorry :(

[github-actions]

Size increase report for "nrfconnect-example-build" from 805a4b3

File	Section	File	VM
chip-shell.elf	device_handles	12	12
chip-shell.elf	text	-764	-764
chip-lighting.elf	device_handles	8	8
chip-lighting.elf	text	-1112	-1112
chip-lock.elf	device_handles	8	8
chip-lock.elf	text	-1112	-1112

Full report output

BLOAT REPORT

Files found only in the build output:
    report.csv

Comparing ./master_artifact/chip-shell.elf and ./pull_artifact/chip-shell.elf:

sections,vmsize,filesize
.strtab,0,34
device_handles,12,12
.shstrtab,0,2
.debug_str,0,-3
.debug_aranges,0,-24
.debug_frame,0,-92
.symtab,0,-112
.debug_abbrev,0,-388
text,-764,-764
.debug_ranges,0,-1480
.debug_line,0,-3120
.debug_loc,0,-4649
.debug_info,0,-7556

Comparing ./master_artifact/chip-lighting.elf and ./pull_artifact/chip-lighting.elf:

sections,vmsize,filesize
device_handles,8,8
.debug_aranges,0,-56
.debug_str,0,-115
.strtab,0,-116
.debug_frame,0,-120
.symtab,0,-336
.debug_abbrev,0,-489
text,-1112,-1112
.debug_ranges,0,-1208
.debug_line,0,-3956
.debug_loc,0,-4189
.debug_info,0,-9883

Comparing ./master_artifact/chip-lock.elf and ./pull_artifact/chip-lock.elf:

sections,vmsize,filesize
device_handles,8,8
.debug_aranges,0,-56
.debug_str,0,-115
.strtab,0,-116
.debug_frame,0,-120
.symtab,0,-336
.debug_abbrev,0,-489
text,-1112,-1112
.debug_ranges,0,-1208
.debug_line,0,-3957
.debug_loc,0,-4197
.debug_info,0,-9858

[bzbarsky-apple]

@kpschoedel Just FYI here. Given the .text savings I guess we should do it, but this is making a pretty major behavior change to a lot of functions (which used to actually null out the buffer pointer in the caller and now don't, or do but only sometimes). @kghost what auditing did you do to make sure all the callers are OK with that behavior change, and what safeguards to we have in place against testing only situations when the pointer is not nulled out and later then ending up in a case when it is?

[bzbarsky-apple]

testing only situations when the pointer is not nulled out and later then ending up in a case when it is

And to be clear, that is exactly where we are after this PR: a bunch of test shims have the "don't null it out" behavior, while their real-code counterparts would probably null it out.