Fix LFN lookup - do not skip LFN entries longer than what we can display

[eduard-sukharev]

Description

Files and directories with LFN longer than configured display limit are shown truncated to that limit. However, when selecting those files by LFN using M23 command, they are skipped as being presumably too long.

This PR fixes that - when detecting whether a entry is LFN we no longer limit LFN sequences to our limits, but consider VFAT specs that allow filenames up to 255 chars (which translates to 255 / 13 = 20 sequences).

Benefits

Selecting filenames and directories longer than 26 chars

Configurations

#define UTF_FILENAME_SUPPORT
#define LONG_FILENAME_HOST_SUPPORT
#define SDSUPPORT

[eduard-sukharev]

Provided that we have files:

/directory_with_very_long_name_1/filename_1.gcode
/directory_with_very_long_name_2/filename_2.gcode

here are the results of following test cases:

Command	Command result	Test result
M23 /directory_with_very_long_name_1/filename_1.gcode	selected filename_1.gcode	✔️
M23 /directory_with_very_long_name_2/filename_2.gcode	selected filename_2.gcode	✔️
M23 /directory_with_very_long_n/filename_1.gcode	selected filename_1.gcode	❌
M23 /directory_with_very_long_n/filename_2.gcode	selected filename_2.gcode	❌

[thinkyhead]

Note that VFAT_ENTRIES_LIMIT is related to the number of characters that can be displayed on LCD and the methods that return the truncated filename to be displayed on LCD. So it should not necessarily be applied to the selection of files by long pathname with M23 or to M20 where we can use a smaller buffer to do the character comparisons. Once we have the selected file's ID and/or short pathname/filename, that is the only piece of data needed to refer to the selected file from that point onward, where we only need to store the complete long name of the selected file for reporting and display.

Meanwhile, the size of the buffer used by SDCARD_SORT_ALPHA is determined by SDSORT_CACHE_VFATS, where we might even store the sorted names in RAM. That might be optimized by just storing the sorted names in one contiguous buffer with a separate index into the buffer for each filename, rather than the current scheme of storing an index into an array of filenames (which mirrors the index into the directory structure used for the non-cached-filename sorting index).