nanocoap: Make coap_block_finish more resilient

[bergzand]

Contribution description

This API change removes the error prone nature of coap_block_finish and
related functions. Before this change the position of the block-wise
options in the coap packet buffer was not allowed to change or the
coap_block_finish function would update the wrong byte in the packet
buffer.

After this change the block option is looked up dynamically from the
packet buffer, ensuring that if the position got changed the function
would still behave correctly.

One case where this would cause issues is when an option got removed via coap_opt_remove(), shifting the block1/2 option.

Testing procedure

Testing the affected examples should be sufficient here

Issues/PRs references

None

Happy 20000 you all! 🎉

[kaspar030]

Happy 20000 everyone! Incredible! 😄

[benpicco]

examples/nanocoap_server does not like this when I run

aiocoap-client -m GET coap://[fe80::748f:e6ff:fed7:426d%tapbr0]/.well-known/core

With this patch:

main(): This is RIOT! (Version: 2024.01-devel-31-g80dbf-HEAD)
RIOT nanocoap example application
Waiting for address autoconfiguration...
{"IPv6 addresses": ["fe80::748f:e6ff:fed7:426d"]}
make: *** [/home/benpicco/dev/RIOT/examples/nanocoap_server/../../Makefile.include:886: term] Segmentation fault (core dumped)

On master:

main(): This is RIOT! (Version: 2024.01-devel)
RIOT nanocoap example application
Waiting for address autoconfiguration...
{"IPv6 addresses": ["fe80::748f:e6ff:fed7:426d"]}

</vfs>,</sha256>,</riot/ver>,</riot/value>,</riot/board>,</echo/>,</.well-known/core>

[bergzand]

Thanks for testing. This PR breaks the "Options Write Buffer API" functions for nanocoap block wise. Not sure how to solve that here.