Handle dot-prefixed names in custom translation

[MichaelChirico]

Closes #1529

• For the specific issue translate_sql() fails when the target for translation starts with a dot #1529, the key IINM is the ls() calls in the names.sql_variant() method.
• I peppered ls_all() everywhere I could see it, without too careful a reading of whether it may not be appropriate in that case. I fear I don't have context to discriminate which should continue being plain ls().

I could go for an assist on writing a regression test, I couldn't quite put all the pieces together. How do I link a custom translation to a simulated connection?

Also LMK if you want more tests covering the switch to ls_all() for the other cases.

[MichaelChirico]

btw, this is not ~quite working, and I'm not sure why -- after this change, the translation looks like (from memory):

translate_sql(.POSIXct(0), con=con)
# <SQL> 1970-01-01 00:00:00

i.e., it's actively evaluated .POSIXct(0) instead of translating it as intended. That's a different behavior than before ("no translation found --> attempt to pass it through to SQL directly"), but not correct.