Yes, accelerators did exist - but they were usually niche products, or handled completely differently (see below).
It's more of a market-driven issue than ability to speed up.
Home computers never really had a big need for speed improvement. After all, any speed up would not only break games, but also be rather expensive, as the host system wasn't really made to be extended in a fundamental way (*1).
More so, manufacturers didn't have much interest in selling a CPU upgrade, when they could do as well by selling a new, faster machine instead. Cases were accelerators were successful are rather exceptions.
The Apple II because, its incarnations didn't offer any speedup for almost 10 years (until the IIgs and IIc+). This rare combination of a huge user base in a rather professional area, plus a manufacturer not upgrading the system, opened a huge window of opportunity for third party offerings of accelerators. Users taking this road usually had a professional need for higher speed -like crunching ever increasing VisiCalc sheets (*2). Much the same way as this group needed larger memory, making a fortune for Saturn Systems. These users were a quite large percentage on the Apple II (*3).
The C64 gives the perfect counter-example. Despite its quite high distribution, only a small minority saw any professional use. As a result, neither Commodore supplied memory expansions (17xx REU), nor did the Super CPU find much use.
The BBC Micro is again an exception, as it was intentionally designed to be expandable. Adding subsystems with other CPUs was intended. Doing the same with a faster 6502 came naturally.
The fact that all of them are 6502 systems is rather by accident.
Z80 systems were basically only used in two large non-intersecting groups: S100-based professional systems and home computers, with MSX being eventually the biggest share.
On the home computer side, manufacturers had no interest in upgrade sales, but unlike Apple, they did offer compatible follow-up models. It's interesting that in this segment it was much more important to add capabilities like better video than speeding up the CPU.
On the professional side, Z80 systems were usually S100 based, enabling easy swap of the CPU board for a new, faster one. Interestingly, the same worked for CP/M on the Apple, where the basic 2 MHz Z80 cards soon were swapped for boards with local memory and Z80 up to 12 MHz.
A special group to be mentioned as well are dedicated (office) systems. Here speed was as well of no concern in itself, but all about the ability to keep up with the user. The Amstrad PCW is a great example here as it sold quite well for more than a decade (1985-1996), making it one of the longest selling machines. A 4 MHz Z80 is more than capable of keeping up with typing text, handling addresses and printing (*4).
*1 - Mostly due the rather tight coupling of CPU and memory/I/O timing.
*2 - We all know how users in management and marketing bloat their spreadsheets way past a point were any rational user would build a custom software, don't we?
*3 - By 1984, less than 1.5 million Apple II were sold - and 700,000 copies of VisiCalc.
*4 - In some way the PCW is the antithesis to the Apple II. Whereas professional Apple II users were at the forefront, willing to spend a lot of money to get top-end tasks done, PCW users were rather the 'working' folks, needing a reliable and performing tool for work that didn't change much.