Make appending operator more efficient

[Witiko]

In #435, we added support for appending to token renderers and renderer prototypes. However, the current implementation requires an extra token list for each token renderer and renderer prototype that stores a copy of the current definition. This is error-prone, since it silently discards any changes to token renderers and renderer prototypes from outside the \markdownSetup command. Furthermore, it is also unnecessary, since we can append directly to the replacement text of token renderers and renderer prototypes as long as the commands are simple unprotected \defs with undelimited parameters, as shown on TeX StackExchange by @eg9, @gucci-on-fleek, @Skillmon and me.

Tasks

• Define function \@@_tl_set_from_cs:NNn following the discussion on TeX StackExchange.
• Remove the extra token list used in appending to token renderers and renderer prototypes.
• Remove error messages failed-to-append-renderer and failed-to-append-renderer-prototype.
• Acknowledge @eg9 and @gucci-on-fleek in CHANGES.md.
• Apply @Skillmon's suggestions and acknowledge @Skillmon in CHANGES.md.
• Document function \@@_tl_set_from_cs:NNn in the technical documentation.
• Use the appending operator in the testing framework.

[Witiko]

If time permits, we may also want to add the function \tl_set_from_cs:NN and its variants discussed on TeX StackExchange to the l3tl module of latex3/latex3 and open a PR. If merged, we should then mark our function \@@_tl_set_from_cs:NNn for removal after we have ceased to support TeX Live 2023.